SECRYPT 2025 Abstracts

Full Papers

Paper Nr:	22
Title:	PPVFL-SplitNN: Privacy-Preserving Vertical Federated Learning with Split Neural Networks for Distributed Patient Data
Authors:	Bashair Alrashed, Priyadarsi Nanda, Hoang Dinh, Amani Aldahiri, Hadeel Alhosaini and Nojood Alghamdi
Abstract:	Medical data privacy regulations pose significant challenges for sharing raw data between healthcare institutions. These challenges are particularly critical when the data is vertically partitioned. In such scenarios, each healthcare provider holds unique but complementary patient information. This makes collaborative learning challenging while protecting patient privacy. As a result, developing effective machine learning models that require integrated data becomes unfeasible. This leads to fragmented analyses and less effective patient care. To address this issue, we developed a vertical federated learning framework using split neural networks to enable secure collaboration while preserving privacy. The framework comprises three main stages: generating symmetric keys to establish secure communication, aligning overlapping patient records across institutions using a privacy-preserving record linkage algorithm, and collaboratively training a global machine learning model without revealing patient privacy. We evaluated the framework on three well-known medical datasets. Our evaluation focused on two critical scenarios: varying degrees of overlap in patient records and differing feature distributions. The proposed framework ensures patient privacy and compliance with strict regulations, providing a scalable and practical solution for real-world healthcare networks. It effectively addresses key challenges in privacy-preserving collaborative machine learning.

Paper Nr:	26
Title:	Evaluating a Bimodal User Verification Robustness Against Synthetic Data Attacks
Authors:	Sandeep Gupta, Rajesh Kumar, Kiran Raja, Bruno Crispo and Carsten Maple
Abstract:	Smartphones balance security and convenience by offering both knowledge-based (PINs, patterns) and biometric (facial, fingerprint) verification methods. However, studies have reported that PINs and patterns can be readily circumvented, while synthetically manipulated face data can easily deceive smartphone facial verification mechanisms. In this paper, we design a bimodal user verification mechanism that combines behavioral (pickup gesture) and biological (face) biometrics for user verification on smartphones. This work establishes a baseline for single-user verification scenarios on smartphones using a one-class verification model. The evaluation is performed in two stages: first, performance is assessed in both unimodal and bimodal settings using publicly available datasets; second, the robustness of the employed biological and behavioral traits is examined against four diverse attacks. Our findings emphasize the necessity of investigating diverse attack vectors, particularly fully synthetic data, to design robust user verification mechanisms.

Paper Nr:	30
Title:	FunBic-CCA: Function Secret Sharing for Biclusterings Applied to Cheng and Church Algorithm
Authors:	Shokofeh VahidianSadegh, Alberto Ibarrondo and Lena Wiese
Abstract:	High-throughput technologies (e.g., the microarray) have fostered the rapid growth of gene expression data collection. These biomedical datasets, increasingly distributed among research institutes and hospitals, fuel various machine learning applications such as anomaly detection, prediction or clustering. In particular, unsupervised classification techniques based on biclustering like the Cheng and Church Algorithm (CCA) have proven to adapt particularly well to gene expression data. However, biomedical data is highly sensitive, hence its sharing across multiple entities introduces privacy and security concerns, with an ever-present threat of accidental disclosure or leakage of private patient information. To address such threat, this work introduces a novel, highly efficient privacy-preserving protocol based on secure multiparty computation (MPC) between two servers to compute CCA. Our protocol performs operations relying on an additive secret sharing and function secret sharing, leading us to reformulate the steps of the CCA into MPC-friendly equivalents. Leveraging lightweight cryptographic primitives, our new technique named FunBic-CCA is first to exploit the efficiency of function secret sharing to achieve fast evaluation of the CCA biclustering algorithm.

Paper Nr:	32
Title:	Honorific Security: Efficient Two-Party Computation with Offloaded Arbitration and Public Verifiability
Authors:	Tianxiang Dai, Yufan Jiang, Yong Li, Jörn Müller-Quade and Andy Rupp
Abstract:	In the secure two-party computation (2PC), an adversary is often categorized as semi-honest or malicious, depending on whether it follows the protocol specifications. Covert security (Aumann and Lindell, 2010) first looks into the “middle ground”, such that an active adversary who cheats will be caught with a predefined probability. Other security notions, such as publicly auditable security (Baum et al., 2014) and (robust) accountability family (Küsters et al., 2010; Graf et al., 2023; Rivinius et al., 2022), achieve public verifiability as a stronger security guarantee by relying on heavy offline and online constructions with zero knowledge proofs and (or) a bulletin board functionality. In this work, we propose a new security notion called honorific security, where an external arbiter can identify the cheater without a bulletin board. Specifically, we delay and outsource the verification steps to the arbiter, so that the original online computation is thus accelerated. We show that a maliciously secure garbled circuit (GC) (Yao, 1986) protocol can be constructed with only slightly more overhead than a passively secure protocol. Our construction performs up to 2.37 times and 13.30 times as fast as the state-of-the-art protocols with covert and malicious security, respectively.

Paper Nr:	36
Title:	A Resilient Randomization Technique Against ML-Driven Timing Side Channel Attack on Mobile Location
Authors:	Abdeslam El-Yahyaoui and Mohammed Erradi
Abstract:	Delivery status notifications are a standard feature of mobile instant messaging applications. They inform users about the successful delivery of their sent messages. However, this common feature opens up a timing side channel attack compromising user location privacy. This attack exploits variations in Round Trip Times (RTTs) across locations, allowing the training of machine learning models for location inference. Recent work proposed a solution based on randomly delaying the RTTs (RDR) on the messenger server side using uniformly sampled perturbations between 0 and a maximum value. I this work, we have shown that the timing side channel attack still persists with significant accuracy even with the aforementioned randomly delaying RTT countermeasure. We then propose a resilient client side randomization technique involving a distribution with randomly varying parameters across RTTs (RVPR). We have shown that the suggested approach (RVPR) is resilient against this attack and has less impact on user experience than the existing RDR approach.

Paper Nr:	43
Title:	Correlation Power Analysis on Ascon with Multi-Bit Selection Function
Authors:	Viet Sang Nguyen, Vincent Grosso and Pierre-Louis Cayrel
Abstract:	Ascon has recently been selected by NIST as the new standard for lightweight cryptography. This highlights the need to evaluate its resilience against implementation attacks such as Correlation Power Analysis (CPA). Traditional CPA on Ascon uses a 1-bit selection function, modeling power consumption based on a single bit of an machine word. However, actual power leakage depends on the entire word. Therefore, the hypothesized power consumption aligns better with the measured values when more bits of the word are involved in the selection function. This paper investigates the use of multi-bit selection functions in CPA on Ascon. We show that the bitsliced-oriented design of Ascon leads the multi-bit selection functions to produce a group of key candidates with high correlations, rather than a single candidate as typically expected in CPA. Through theoretical analysis and experimental validation, we examine this behavior in detail. Based on these insights, we propose an efficient key recovery algorithm tailored for the multi-bit selection functions. Our results demonstrate that this approach significantly reduces the number of CPA runs required for full key recovery.

Paper Nr:	49
Title:	Lens Aberrations Detection and Digital Camera Identification with Convolutional Autoencoders
Authors:	Jarosław Bernacki and Rafał Scherer
Abstract:	Digital camera forensics relies on the ability to identify digital cameras based on their unique characteristics. While many methods exist for camera fingerprinting, they often struggle with efficiency and scalability due to the large image sizes produced by modern devices. In this paper, we propose a novel approach that utilizes convolutional and variational autoencoders to detect optical aberrations, such as vignetting and distortion. Our model, trained in an aberration-independent manner, enables automatic detection of these distortions without needing reference patterns. Furthermore, we demonstrate that the same methodology can be applied to digital camera identification based on image analysis. Extensive experiments conducted on multiple cameras and images confirm the effectiveness of our approach in both aberration detection and device fingerprinting, highlighting its potential applications in forensic investigations.

Paper Nr:	56
Title:	RingAuth: User Authentication Using a Smart Ring
Authors:	Jack Sturgess, Simon Birnbach, Simon Eberz and Ivan Martinovic
Abstract:	We show that by using inertial sensor data generated by a smart ring, worn on the finger, the user can be authenticated when making mobile payments or when knocking on a door (for access control purposes). We also demonstrate that smart ring data can authenticate payments made with a smartwatch, and vice versa, such that either device can act as an implicit second factor for the other when worn on the same arm. To validate the system, we conducted a user study (n=21) to collect finger and wrist motion data from users as they perform gestures, and we evaluate the system against an active impersonation attacker. We develop payment authentication and access control models for which we achieve equal error rates of 0.04 and 0.02, respectively.

Paper Nr:	62
Title:	Post-Quantum Secure Channel Protocols for eSIMs: Design, Validation and Performance Analysis
Authors:	Luk Bettale, Emmanuelle Dottax and Laurent Grémy
Abstract:	The transition to Post-Quantum (PQ) cryptography is increasingly mandated by national agencies and organizations, often involving a phase where classical and PQ primitives are combined into hybrid solutions. In this context, existing protocols must be adapted to ensure quantum resistance while maintaining their security goals. These adaptations can significantly impact performance, particularly on embedded devices. In this article, we focus on standardized protocols which support application management on eSIMs across different modes. This is a complex use-case, involving constrained devices with stringent security requirements. We present PQ adaptations, including both hybrid and fully PQ versions, for all modes. Using ProVerif, we provide automated proofs that verify the security of these PQ variants. Additionally, we analyze the performance impact of implementing PQ protocols on devices, measuring runtime and bandwidth consumption. Our findings highlight the resource overhead associated with achieving post-quantum security for eSIM management.

Paper Nr:	70
Title:	Bolstering IIoT Resilience: The Synergy of Blockchain and CapBAC
Authors:	Argiro Anagnostopoulou, Eleni Kehrioti, Ioannis Mavridis and Dimitris Gritzalis
Abstract:	The growing integration of Internet of Things (IoT) into industrial environments highlights the need for adequate security and privacy maintenance. While traditional access control methods fall short in addressing the rising challenges of such environments, the combination of capability-based access control (CapBAC) models with blockchain technology emerges as a promising alternative. In this paper, we conduct a comprehensive analysis and comparison of approaches that integrate these two concepts. The evaluation of each approach is based on twelve criteria, including scalability, performance, efficiency, latency, throughput, degree of decentralization, consensus mechanism, smart contracts adoption, complexity, interoperability, security guarantees, and privacy. The aim of our analysis is to examine whether the combination of CapBAC and Blockchain brings a new era of secure industrial IoT (IIoT) operations. In order to identify the strengths and the areas for improvement, we provide four types of comparison to further assess these approaches based on IIoT requirements. Finally, we thoroughly discuss our findings, indicating directions for future research in order to enhance the adoption of such innovative mechanisms across broader industrial landscapes.

Paper Nr:	75
Title:	Privacy2Practice: Leveraging Automated Analysis for Privacy Policy Transparency and Compliance
Authors:	Saja Alqurashi and Indrakshi Ray
Abstract:	Privacy policies play a critical role in safeguarding information systems, yet they are frequently expressed in lengthy, complex natural language documents. The intricate and dense language of these policies poses sub-stantial challenges, making it difficult for both novice users and experts to fully comprehend data collection, sharing practices, and the overall transparency of data handling. This issue is particularly concerning given the necessity of disclosing data practices to users, as mandated by privacy regulations such as the General Data Protection Regulation (GDPR) and the California Consumer Privacy Act (CCPA). To address these challenges and improve data transparency, this paper introduces Privacy2Practice, a comprehensive automated framework leveraging Natural Language Processing (NLP) techniques to extract and analyze key information from privacy policies. By automating the identification of data practices mandated by privacy regulations, the framework assesses how transparently these practices are disclosed, ensuring better alignment with regulatory requirements.The proposed approach significantly enhances the transparency and the compliance of privacy policies by identifying entities (F1-scores: 97% for first-party and 93% for third-party entities), data types (F1-score: 82%), and purposes of data collection and sharing (F1-score: 90%). These results underscore the importance of transparency, particularly when data is shared with external parties, and highlight the challenges associated with automating privacy policy analysis. The results highlight significant challenges, such as undisclosed third-party sharing, while showcasing the potential of automation to be more comprehensive, transparent and compliant with regulatory standards.

Paper Nr:	76
Title:	SCAM: Secure Shared Cache Partitioning Scheme to Enhance Throughput of CMPs
Authors:	Varun Venkitaraman, Rishab Ravi, Tejeshwar Bhagatsing Thorawade, Nirmal Kumar Boran and Virendra Singh
Abstract:	Utility-based dynamic cache partitioning scheme (UCP) improves performance in chip multiprocessors (CMPs) by dynamically way-partitioning the shared Last-Level Cache (LLC) based on each core’s utility. At the end of every phase, UCP allocates more ways to the core with higher utility. However, the process of transferring ownership of a cache way from low utility core to high utility core on a cache miss (when there is partition decision change) creates side channels, making shared LLCs vulnerable to data leaks. PASS-P addresses these vulnerabilities by invalidating cache lines before transferring ownership from one core to another after partition change. While it provides security, our analysis shows that PASS-P does not always choose the best cache line for transfer, leaving room for improving performance. To improve performance of the system without compromising on security, we propose SCAM, a secure shared cache partitioning scheme. SCAM optimizes the process of selection of transfer candidates, improving performance over PASS-P while maintaining security. SCAM achieves up to 4% performance improvement over PASS-P and reduces LLC misses per kilo instructions (MPKI) by up to 5%. SCAM offers an efficient solution for secure dynamic way-partitioning in shared caches of multi-core systems. It provides better performance without compromising security, making it an effective strategy for protecting against side-channel attacks while ensuring optimal cache utilization.

Paper Nr:	81
Title:	Weak, Weak-Insider, and Randomized Weak Privacy in the HPVP Model for RFID
Authors:	Ferucio Laurenţiu Ţiplea
Abstract:	RFID schemes that provide weak privacy or similar privacy forms are useful in any domain where the adversary cannot mount a corruption attack. In addition, these schemes can be constructed using only symmetric cryptography and can provide time-efficient identification. This paper focuses on RFID schemes that provide weak privacy in the Hermans-Pashalidis-Vercauteren-Preneel (HPVP) model based on tag indistinguishability. We first show that no adversary can have a non-negligible advantage in distinguishing between keys of a pseudo-random function. We then use this result to highlight RFID schemes that provide weak, weak-insider, and randomized weak privacy in the model above.

Paper Nr:	89
Title:	An NLP-Based Framework Leveraging Email and Multimodal User Data
Authors:	Neda Baghalizadeh-Moghadam, Frédéric Cuppens and Nora Boulahia-Cuppens
Abstract:	Traditional approaches for insider threat detection rely on analyzing activity logs to detect abnormal user activities. In this paper, we investigate how the exchange of messages between users could also contribute to detecting insider threats. This work presents an NLP-driven anomaly detection framework that incorporates feature engineering and prompt engineering across multimodal user activities, such as emails, HTTP requests, file access, and logon events. This study employs Named Entity Recognition (NER), Sentiment Analysis, and Prompt Engineering, to extract semantic, contextual, and behavioral insights that enhance anomaly detection. These enriched representations are processed by an Isolation Forest and One-Class Support Vector Machine (One-Class SVM) for the unsupervised detection of deviations from normal user behavior. Unlike most previous works that focus solely on user log activity datasets, our method incorporates both user log activity and email communication data for insider threat detection. Experimental results on the CERT r4.2 dataset demonstrate that the proposed multimodal approach improves anomaly detection with high accuracy, greater precision, and reduced false alarm rates. Hence, our framework offers greater explainability and scalability in addressing sophisticated insider threats.

Paper Nr:	90
Title:	Beyond Rules: How Large Language Models Are Redefining Cryptographic Misuse Detection
Authors:	Zohaib Masood and Miguel Vargas Martin
Abstract:	The use of Large Language Models (LLMs) in software development is rapidly growing, with developers increasingly relying on these models for coding assistance, including security-critical tasks. Our work presents a comprehensive comparison between traditional static analysis tools for cryptographic API misuse detec-tion—CryptoGuard, CogniCrypt, and Snyk Code—and the LLMs—GPT, Llama, Claude, and Gemini. Using benchmark datasets (OWASP, CryptoAPI, and MASC), we evaluate the effectiveness of each tool in identifying cryptographic misuses. Our findings show that GPT 4-o-mini surpasses current state-of-the-art static analysis tools on the CryptoAPI and MASC datasets, though it lags on the OWASP dataset. Additionally, we assess the quality of LLM responses to determine which models provide actionable and accurate advice, giving developers insights into their practical utility for secure coding. This study highlights the comparative strengths and limitations of static analysis versus LLM-driven approaches, offering valuable insights into the evolving role of AI in advancing software security practices.

Paper Nr:	93
Title:	Did You Break the Glass Properly? A Policy Compliance Framework for Protected Health Information (PHI) Emergency Access
Authors:	Md Al Amin, Rushabh Shah, Hemanth Tummala and Indrajit Ray
Abstract:	HIPAA, HITECH, GDPR, and other data protection laws and regulations mandate patients’ consent to access and share their data. They also impose compliance requirements for healthcare organizations. Non-compliance cases or failure to comply come with financial, reputational, business, and other penalties. In emergency medical situations, accessing a patient’s protected health information or records can be critical to saving lives, especially when the patient is unconscious or unable to consent. This paper addresses the need for a secure, compliant, auditable system for emergency PHI access. We propose a blockchain and smart contract-based policy compliance framework where the emergency duty doctor requests access and must obtain approval from the senior in charge, which is recorded through multi-signature transactions. Once access is granted, the patient or their emergency contact is notified. To prevent unauthorized modifications, all actions are captured as immutable audit logs within a private blockchain network. The compliance check uses a novel Proof of Compliance (PoC) consensus mechanism, ensuring all access requests adhere to defined policies. This framework offers transparency, accountability, and security for emergency PHI access requirements.

Paper Nr:	103
Title:	Empirical Evaluation of Memory-Erasure Protocols
Authors:	Reynaldo Gil-Pons, Sjouke Mauw and Rolando Trujillo-Rasua
Abstract:	Software-based memory-erasure protocols are two-party communication protocols where a verifier instructs a computational device to erase its memory and send a proof of erasure. They aim at guaranteeing that low-cost IoT devices are free of malware by putting them back into a safe state without requiring secure hardware or physical manipulation of the device. Several software-based memory-erasure protocols have been introduced and theoretically analysed. Yet, many of them have not been tested for their feasibility, performance and security on real devices, which hinders their industry adoption. This article reports on the first empirical analysis of software-based memory-erasure protocols with respect to their security, erasure guarantees, and performance. The experimental setup consists of 3 modern IoT devices with different computational capabilities, 7 protocols, 6 hash-function implementations, and various performance and security criteria. Our results indicate that existing software-based memory-erasure protocols are feasible, although slow devices may take several seconds to erase their memory and generate a proof of erasure. We found that no protocol dominates across all empirical settings, defined by the computational power and memory size of the device, the network speed, and the required level of security. Interestingly, network speed and hidden constants within the protocol specification played a more prominent role in the performance of these protocols than anticipated based on the related literature. We provide an evaluation framework that, given a desired level of security, determines which protocols offer the best trade-off between performance and erasure guarantees.

Paper Nr:	112
Title:	Blockchain-Based Multi-Signature System for Critical Scenarios
Authors:	Cristina Alcaraz, Davide Ferraris, Hector Guzman and Javier Lopez
Abstract:	Blockchain technology plays a crucial role in securing and streamlining transactions across various critical domains. For that reason, this paper presents a Blockchain-based multi-signature system designed for high-stakes scenarios, where both user and Blockchain-generated signatures are required to authorize transactions. By integrating smart contracts, multi-signature coordination, and Blockchain validation, the proposed architecture enhances security, accountability, and resilience. The framework is applied to two key sectors: Mobility and energy. In mobility, it addresses two distinct use cases: Ambulance services, where secure and verifiable authorization of emergency access is required, and insurance claim processing, ensuring transparent, tamper-proof validations. In the energy sector, the system facilitates decentralized, trust-enhanced peer-to-peer energy trading by guaranteeing transaction integrity and compliance. The architecture leverages smart contracts to enforce transaction policies, aggregate multi-signatures, and validate operations while maintaining transparency and reliability. This work highlights the importance of decentralized decision-making and immutable records in securing critical infrastructures. Future research will focus on optimizing performance and evaluating the system’s integration with existing Blockchain platforms such as Ethereum and Hyperledger.

Paper Nr:	129
Title:	Control Flow Protection by Cryptographic Instruction Chaining
Authors:	Shahzad Ahmad, Stefan Rass, Maksim Goman, Manfred Schlägl and Daniel Große
Abstract:	We present a novel secure execution environment that provides comprehensive protection for program execution through a unified cryptographic approach. Our construction employs authenticated encryption, ensuring instruction confidentiality, integrity, and correct execution ordering. The system’s key innovation lies in its cryptographic binding of consecutive instructions through a novel key chaining mechanism that prevents instruction reordering and replay attacks while maintaining an enforced order of instructions using cryptographic chaining via keys. We introduce specialized handling for control flow operations, including branches, jumps, and function calls, that preserves security guarantees across complex execution paths. The framework includes a protection mechanism for registers and memory, creating a fully secured execution environment. Our performance analysis quantifies the computational overhead and provides a Python proof-of-concept implementation that validates the practical viability of our approach https: //github.com/shahzadssg/Control-Flow-Protection-by-Cryptographic-Instruction-Chaining.git.

Paper Nr:	136
Title:	PRIVÉ: Towards Privacy-Preserving Swarm Attestation
Authors:	Nada El Kassem, Wouter Hellemans, Ioannis Siachos, Edlira Dushku, Stefanos Vasileiadis, Dimitrios S. Karas, Liqun Chen, Constantinos Patsakis and Thanassis Giannetsos
Abstract:	In modern large-scale systems comprising multiple heterogeneous devices, the introduction of swarm attestation schemes aims to alleviate the scalability and efficiency issues of traditional single-Prover and single-Verifier attestation. In this paper, we propose PRIV´E , a privacy-preserving, scalable, and accountable swarm attestation scheme that addresses the limitations of existing solutions. Specifically, we eliminate the assumption of a trusted Verifier, which is not always applicable in real-world scenarios, as the need for the devices to share identifiable information with the Verifier may lead to the expansion of the attack landscape. To this end, we have designed an enhanced variant of the Direct Anonymous Attestation (DAA) protocol, offering traceability and linkability whenever needed. This enables PRIV´E to achieve anonymous, privacy-preserving attestation while also providing the capability to trace a failed attestation back to the compromised device. To the best of our knowledge, this paper presents the first Universally Composable (UC) security model for swarm attestation accompanied by mathematical UC security proofs, as well as experimental benchmarking results that highlight the efficiency and scalability of the proposed scheme.

Paper Nr:	151
Title:	An Enhanced Two-Step CPA Side-Channel Analysis Attack on ML-KEM
Authors:	Mark Kennaway, Tuan Hoang, Ayesha Khalid, Ciara Rafferty and Máire O’Neill
Abstract:	This work presents an enhanced two-step Correlation Power Analysis (CPA) attack targeting the recently standardised ML-KEM on an ARM Cortex M4. Our enhancement exploits the knowledge of intermittent variables to identify sample points of interest and develop bespoke attack functions. Step one targets the odd coefficients of each Secret Key Polynomial Vector ( ˆs), before step two targets the remaining even coefficients using more elaborate attack functions. After successfully demonstrating key recovery for the first set of ˆs, we then characterise leakage behaviour, revealing a trend indicating recovery of each coefficient becomes more efficient with subsequent iterations of the internal doublebasemul operation. By applying our enhanced twostep attack methodology, we successfully recovered the entire key using only 179 traces, without the need for elaborate preconditions or ciphertext manipulations. We obtain remarkable results in the initial stage of our attack, while the second phase achieves performance comparable to other recent studies.

Paper Nr:	155
Title:	Can Contributing More Put You at a Higher Leakage Risk? The Relationship Between Shapley Value and Training Data Leakage Risks in Federated Learning
Authors:	Soumia Zohra El Mestari, Maciej Krzysztof Zuziak, Gabriele Lenzini and Salvatore Rinzivillo
Abstract:	Federated Learning (FL) is a crucial approach for training large-scale AI models while preserving data locality, eliminating the need for centralised data storage. In collaborative learning settings, ensuring data quality is essential, and in FL, maintaining privacy requires limiting the knowledge accessible to the central orchestrator, which evaluates and manages client contributions. Accurately measuring and regulating the marginal impact of each client’s contribution needs specialised techniques. This work examines the relationship between one such technique—Shapley Values—and a client’s vulnerability to Membership inference attacks (MIAs). Such a correlation would suggest that the contribution index could reveal high-risk participants, potentially allowing a malicious orchestrator to identify and exploit the most vulnerable clients. Conversely, if no such relationship is found, it would indicate that contribution metrics do not inherently expose information exploitable for powerful privacy attacks. Our empirical analysis in a cross-silo FL setting demonstrates that leveraging contribution metrics in federated environments does not substantially amplify privacy risks.

Paper Nr:	169
Title:	Synthetic and (Un)Secure: Evaluating Generalized Membership Inference Attacks on Image Data
Authors:	Pasquale Coscia, Stefano Ferrari, Vincenzo Piuri and Ayse Salman
Abstract:	Synthetic data are widely employed across diverse fields, including computer vision, robotics, and cybersecurity. However, generative models are prone to unintentionally revealing sensitive information from their training datasets, primarily due to overfitting phenomena. In this context, membership inference attacks (MIAs) have emerged as a significant privacy threat. These attacks employ binary classifiers to verify whether a specific data sample was part of the model’s training set, thereby discriminating between member and non-member samples. Despite their growing relevance, the interpretation of MIA outcomes can be misleading without a detailed understanding of the data domains involved during both model development and evaluation. To bridge this gap, we performed an analysis focused on a particular category (i.e., vehicles) to assess the effectiveness of MIA under scenarios with limited overlap in data distribution. First, we introduce a data selection strategy, based on the Fréchet Coefficient, to filter and curate the evaluation datasets, followed by the execution of membership inference attacks under varying degrees of distributional overlap. Our findings indicate that MIAs are highly effective when the training and evaluation data distributions are well aligned, but their accuracy drops significantly under distribution shifts or when domain knowledge is limited. These results highlight the limitations of current MIA methodologies in reliably assessing privacy risks in generative modeling contexts.

Short Papers

Paper Nr:	14
Title:	Towards Quantum Machine Learning in Ransomware Detection
Authors:	Francesco Mercaldo, Giovanni Ciaramella, Fabio Martinelli and Antonella Santone
Abstract:	Ransomware represent one of the most aggressive malware, due to their capability to prevent access to data and, as a consequence, totally paralyze the activity of any organization, such as companies, but also hospitals or banks. Considering the inadequacy of the signature-based approach, mainly exploited by free and commercial current antimalware, researchers are proposing new ransomware detection techniques based on deep learning. Recently, with the introduction of quantum computing, there is the possibility to introduce quantum principles into machine learning. In this paper, we propose an approach for ransomware detection through a quantum machine learning model aimed to analyse images obtained from the application opcodes. In particular, a hybrid model is proposed, composed of quantum and convolutional layers to discern between ransomware, generic malware, and trusted applications. To demonstrate that quantum machine learning is promising in ransomware detection, a real-world dataset composed by 15,000 applications is evaluated, by showing that the proposed hybrid quantum model obtains promising performances if compared to (fully) convolutional models (i.e., Alex Net, MobileNet, and a convolutional model developed by authors).

Paper Nr:	25
Title:	A Hybrid-Based Transfer Learning Approach for IoT Device Identification
Authors:	Stephanie M. Opoku, Habib Louafi and Malek Mouhoub
Abstract:	The rapid growth and diversity of Internet of Things (IoT) devices pose significant challenges in device identification and network security. Traditional techniques for fingerprinting IoT devices frequently encounter challenges due to the complexity and scale of today’s IoT networks. This paper presents an innovative model applying transfer learning (TL) methodologies to analyze network data and precisely identify IoT devices. Our solution effectively classifies devices by integrating instance-based, feature-based, and hybrid-based TL methodologies for extracting essential features from traffic data. The proposed model undergoes a thorough evaluation on three benchmark IoT datasets, exhibiting improved prediction accuracy, precision, recall, and F1-Score relative to traditional methods. The hybrid technique significantly improves performance by handling computational and scalability issues. This paper highlights the effectiveness of TL in improving IoT device identification, providing an efficient and effective solution for various and dynamic network environments.

Paper Nr:	29
Title:	Insider Threats and Countermeasures Based on AI Lie Detection
Authors:	Konstantinos Kalodanis, Panagiotis Rizomiliotis, Charalampos Papapavlou, Apostolos Skrekas, Stavros Papadimas and Dimosthenis Anagnostopoulos
Abstract:	Insider threats continue to pose some of the most significant security risks within organizations, as malicious insiders have privileged access to sensitive or even classified data and systems. This paper explores an emerging approach that applies Artificial Intelligence (AI)–based lie detection techniques to mitigate insider threats. We investigate state-of-the-art AI methods adapted from Natural Language Processing (NLP), physiological signal analysis, and behavioral analytics to detect deceptive behavior. Our findings suggest that the fusion of multiple data streams, combined with advanced AI classifiers such as transformer-based models and Graph Neural Networks (GNN), leads to enhanced lie detection accuracy. Such systems must be designed in accordance with EU AI Act, which imposes requirements on transparency, risk management, and compliance for high-risk AI systems. Experimental evaluations on both synthesized and real-world insider threat datasets indicate that the proposed methodology achieves a performance improvement of up to 15–20% over conventional rule-based solutions. The paper concludes by exploring deployment strategies, limitations, and future research directions to ensure that AI-based lie detection can effectively and ethically bolster insider threat defences.

Paper Nr:	37
Title:	On the Security of Opportunistic Re-Keying
Authors:	Stefan Lucks, David Schatz and Guenter Schaefer
Abstract:	Asymmetric cryptography is a cornerstone for security in modern IT infrastructures like virtual private networks (VPNs). Unfortunately, the security of currently deployed schemes is threatened by the ongoing research in quantum computing. And while quantum-resistant alternatives exist, known as post-quantum cryptography (PQC), analyses regarding their (implementation) security are not as mature, yet. Consequently, solely relying on PQC might be susceptible to “store now, decrypt later” attacks. Instead, many researchers suggest using “hybrid” key exchanges, e.g., combining classical asymmetric cryptography, PQC, and symmetric alternatives like quantum key distribution (QKD) and multipath key reinforcement (MKR). In this article, we formalize the idea of “opportunistic re-keying”, where a session key is continuously updated using input key material that might be known or even chosen by an attacker. Assuming that at least one input key material is not known to the attacker, we prove the security of the construction in the random oracle model. I.e., when an ideal random function is used for combining the current internal state and new input to generate the next session key and state. Further, we suggest two concrete parameter sets for the construction, corresponding to the security categories 3 and 5 of the NIST standardization process for PQC.

Paper Nr:	38
Title:	Secure and Practical Cold (and Hot) Staking
Authors:	Mario Larangeira
Abstract:	The stake delegation technique is what turns the general Proof of Stake (PoS) into a practical protocol for a large number of participants, ensuring the security of the distributed system, in what is known as Delegated PoS (DPoS). Karakostas et al. (SCN ’20) formalized the delegation method paving the way for a whole industry of stake pools by proposing a formal definition for wallet as a universal composable (UC) functionality and introducing a corresponding protocol. On the other hand, a widely used technique named hot/cold wallet was formally studied by Das et al. (CCS ’19 and ’21), and Groth and Shoup (Eurocrypt ’22) for different key derivation methods in the Proof of Work (PoW) setting, but not PoS. Briefly, while hot wallets are exposed to the risks of the network, the cold wallet is kept offline, thus more secure. However this may impair some capabilities given that the cold wallet is kept indefinitely offline. It is straightforward to observe that this “double wallet” design is not naturally portable to the setting where delegation is paramount, i.e., DPoS. This work identifies challenges for PoS Hot/Cold Wallet and proposes a secure and practical protocol.

Paper Nr:	40
Title:	Privacy-Preserving Machine Learning in IoT: A Study of Data Obfuscation Methods
Authors:	Yonan Yonan, Mohammad O. Abdullah, Felix Nilsson, Mahdi Fazeli, Ahmad Patooghy and Slawomir Nowaczyk
Abstract:	In today’s interconnected digital world, ensuring data privacy is critical, particularly for neural networks operating remotely in the age of the Internet of Things (IoT). This paper tackles the challenge of data privacy preservation in IoT environments by investigating Utility-Preserving Data Transformation (UPDT) methods, which aim to transform data in ways that reduce or eliminate sensitive information while retaining its utility for analytical tasks. UPDT methods aim to balance privacy preservation and utility in data analytics. This study examines the strengths and limitations of these methods, focusing on ObfNet, a neural network-based obfuscation algorithm, as a representative case study to contextualize our analysis. By analyzing ObfNet, we highlight its vulnerabilities and based on these findings we introduce LightNet and DenseNet as novel neural networks to identify ObfNet’s limitations, particularly for larger and more complex data. We uncover challenges such as information leakage and explore the implications for maintaining privacy during remote neural network inference. Our findings underscore the challenges and possibilities to preserve privacy during remote neural network inference for UPDT algorithms, especially in resource-limited edge devices.

Paper Nr:	57
Title:	Addressing the C/C++ Vulnerability Datasets Limitation: The Good, the Bad and the Ugly
Authors:	Claudio Curto, Daniela Giordano and Daniel Gustav Indelicato
Abstract:	Recent years have witnessed growing interest in applying deep learning techniques to software security assessment, particularly for detecting vulnerability patterns in human-generated source code. Despite advances, the effectiveness of deep learning models is often hindered by limitations in the datasets used for training. This study conducts a comprehensive evaluation of one widely used and two recently released C/C++ real-world vulnerable code datasets to assess their impact on the performance of transformer-based models, focusing on generalization across unseen projects, unseen vulnerability types and diverse data distributions. In addition, we analyze the effects of aggregating datasets and compare the results with previous experiments. Experimental results demonstrate that combining datasets significantly improves model generalization across varied distributions, highlighting the importance of diverse, high-quality data for enhancing vulnerability detection in source code.

Paper Nr:	59
Title:	Large Language Models as Carriers of Hidden Messages
Authors:	Jakub Hoscilowicz, Pawel Popiolek, Jan Rudkowski, Jedrzej Bieniasz and Artur Janicki
Abstract:	Simple fine-tuning can embed hidden text into large language models (LLMs), which is revealed only when triggered by a specific query. Applications include LLM fingerprinting, where a unique identifier is embedded to verify licensing compliance, and steganography, where the LLM carries hidden messages disclosed through a trigger query. Our work demonstrates that embedding hidden text via fine-tuning, although seemingly secure due to the vast number of potential triggers, is vulnerable to extraction through analysis of the LLM’s output decoding process. We introduce an extraction attack called Unconditional Token Forcing (UTF), which iteratively feeds tokens from the LLM’s vocabulary to reveal sequences with high token probabilities, indicating hidden text candidates. We also present Unconditional Token Forcing Confusion (UTFC), a defense paradigm that makes hidden text resistant to all known extraction attacks without degrading the general performance of LLMs compared to standard fine-tuning. UTFC has both benign (improving LLM fingerprinting) and malign applications (using LLMs to create covert communication channels).

Paper Nr:	71
Title:	TIGER: TrIaGing KEy Refreshing Frequency via Digital Sensors
Authors:	Md Toufiq Hasan Anik, Hasin Ishraq Reefat, Mohammad Ebrahimabadi, Javad Bahrami, Hossein Pourmehrani, Jean-Luc Danger, Sylvain Guilley and Naghmeh Karimi
Abstract:	Key refreshing is a pragmatic countermeasure to side-channel attacks, designed to revoke and replace the key promptly when an attack is either anticipated or suspected. This system-level approach rekeys the device under attack and keeps paired devices protected if cryptography secures data in transit. The frequency of key refreshing is critical: too infrequent, and security risks increase; too frequent, and system performance degrades. This frequency is set pre-silicon via threat analysis but may be inefficient as leakage varies with operating conditions. To fill the gap, we introduce a post-silicon strategy for optimal rekeying frequency. Our proposed scheme TIGER deploys a digital sensor to monitor environmental conditions and enabling inference at runtime by pre-characterizing the leakage rate correlated to the digital sensor measurements. The accumulated leakage rate helps deduce a cutoff time for rekeying. We demonstrate the end-to-end feasibility of this approach on an FPGA board designed for side-channel threat assessment.

Paper Nr:	72
Title:	MICODE: A Minimal Code Design for Secret Sharing Scheme
Authors:	Belkacem Imine, Rahul Saha and Mauro Conti
Abstract:	Secret Sharing Schemes (SSS) in cryptography often utilize minimal linear codes for efficiency, with minimum distance playing a crucial role. Determining minimal codewords in general linear codes presents a challenge known as the linear code covering problem. To address this, we propose MInimal COde DEsign (MICODE), a novel method for generating minimal codes from binary Reed-Muller (RM) codes using the Ashikhmin-Barg lemma. Unlike existing approaches limited to small RM codes, MICODE extends to higher orders through a systematic puncturing strategy. By recursively removing one-weight columns from the generator matrix, we reduce the RM code’s maximum Hamming weight preserving its minimum distance. The RM generator matrix’s structure, derived from the Kronecker product of lower-triangular binary matrices, facilitates this construction. We conduct rigorous mathematical analysis of MICODE establishing parameters for a secure SSS. While these minimal codes are unsuitable for error correction due to their reduced code rate, they are proven highly effective for cryptographic applications, such as Massey SSS, where security depends on minimum distance. Our analysis also explores trade-offs between code rate and error performance offering new insights into their theoretical and practical implications.

Paper Nr:	78
Title:	I Know What You Bought Last Summer: Investigating User Data Leakage in E-Commerce Platforms
Authors:	Ioannis Vlachogiannakis, Emmanouil Papadogiannakis, Panagiotis Papadopoulos, Nicolas Kourtellis and Evangelos Markatos
Abstract:	In the digital age, e-commerce has transformed the way consumers shop, offering convenience and accessibility. Nevertheless, concerns about the privacy and security of personal information shared on these platforms have risen. In this work, we investigate user privacy violations, noting the risks of data leakage to third-party entities. Utilizing a semi-automated data collection approach, we examine a selection of popular online e-shops, revealing that nearly 30% of them violate user privacy by disclosing personal information to third parties. We unveil how minimal user interaction across multiple e-commerce websites can result in a comprehensive privacy breach. We observe significant data-sharing patterns with platforms like Facebook, which use personal information to build user profiles and link them to social media accounts.

Paper Nr:	84
Title:	RAHE: A Robust Attribute-Based Aggregate Scheme Enhanced with Homomorphic Encryption for 5G-Connected Delivery Drones
Authors:	Aagii Mariam Thomas and Sana Belguith
Abstract:	Unmanned Aerial Vehicles (UAVs), commonly known as drones, have become essential for transporting packages, food, medicines, and other goods due to the growing demand for fast and efficient delivery services. The implementation of 5G technology provides high-speed, low-latency, and reliable connectivity, which allows drones to exchange mission-critical data effectively. However, drones utilizing 5G networks are susceptible to security threats that could compromise essential security requirements such as confidentiality, authentication, integrity, and availability. In this paper, we propose a robust communication framework designed for secure interactions among 5G-connected delivery drones. Our framework relies on a novel Attribute-Based Encryption with Aggregation that is composed by an enhanced multi-level Attribute-Based Encryption (ABE) scheme with Homomorphic Encryption (HE). By integrating HE with the ABE scheme, the Ground Control Station (GCS) and the parent drone can decrypt mission-critical messages as required. This ensures that only authorized entities have access to sensitive data. Additionally, in scenarios that require data aggregation without exposing the underlying content, the HE property within the ABE scheme facilitates this process. As a result, encrypted data subsets can be aggregated anywhere in the network without the need for decryption, thereby preserving data confidentiality and enhancing both communication and computational efficiency. We utilize a hierarchical Chain-Based Data Aggregation (CBDA) model for the structural organization of drones, which enhances communication efficiency and reduces energy consumption. By integrating multi-level ABE for flexible and secure access control with HE, our framework effectively addresses major security challenges faced by 5G-based drone networks, ensuring the security and efficient management of mission-critical data.

Paper Nr:	91
Title:	Secure and Hybrid Clustering for IoT Networks: An Adaptive Dynamic Reconfigurability Approach
Authors:	Osama Mohammed Dighriri, Priyadarsi Nanda, Manoranjan Mohanty and Ibrahim Haddadi
Abstract:	Current Internet of Things (IoT) networks face significant challenges in scalability, energy efficiency, and security within resource-constrained environments. This paper proposes a hybrid clustering framework combining BIRCH with DBSCAN algorithms while integrating AES-128 encryption for secure communication. Our proposed scheme is implemented using Contiki-NG simulator and analyzed using Python 3. Our approach demonstrates a 20% improvement in resource utilization, 43.26% reduction in latency, and 99.91% message success rate conducted across 2,154 test runs, with security overhead below 5%. This reduces cluster reconfiguration frequency and improves system stability, addressing limitations in adaptability, data integrity, and optimization for dynamic IoT infrastructures.

Paper Nr:	92
Title:	VKG2AG : Generating Automated Knowledge-Enriched Attack Graph (AG) from Vulnerability Knowledge Graph (VKG)
Authors:	Md Rakibul Hasan Talukder, Rakesh Podder and Indrajit Ray
Abstract:	Attack Graph (AG) analysis is a well-established technique to asses security threats in networked systems. However, traditional AGs primarily rely on coarse level vulnerability information from the Common Vulnerabilities and Exposures (CVE) repository for identifying attack paths and suggesting patch-based mitigation strategies. This approach presents significant limitations, including unavailability of patches, compatibility constraints, and system downtime, leaving security analysts without viable alternatives for optimized risk mitigation. To address this challenge, we propose two new paradigms: a novel knowledge-enriched AG framework and a Vulnerability Knowledge Graph (VKG). VKG incorporate fine-grained, structured vulnerability information that allows exploration of additional attack mitigation strategies beyond vulnerability patching in the AG analysis. We formally define VKG and AG along with algorithms for automated knowledge build-up, integration, and querying. To ensure seamless interoperability, we develop an interface that facilitates dynamic knowledge transfer between VKG and AG, enabling enhanced security reasoning without introducing inter-dependencies. We evaluate our methodology on a test network and demonstrate how the knowledge-driven AG can improve security decision-making by providing system administrators with adaptable, scenario-based defense mechanisms with actionable insights.

Paper Nr:	96
Title:	Learning Personalized and Context-Aware Violation Detection Rules in Trigger-Action Apps
Authors:	Mahsa Saeidi, Sai Sree Laya Chukkapalli, Anita Sarma and Rakesh B. Bobba
Abstract:	Trigger-action apps are being increasingly used by end users to connect smart devices and online services to create new functionality. However, these apps can cause undesirable implicit information flows (secrecy violation) or lead to unintended accesses (integrity violation) depending on the usage context. Existing solutions designed to address such risks rely on predefined rules to control and mitigate such implicit information flows or unintended accesses. However, defining such rules is difficult for end users. In this work, we propose a learning-based approach to learn rules that flag violating situations based on the usage context. We also propose a set of reduction steps to reduce the complexity of the learned rules. We are able to achieve a good F1-measure in predicting both secrecy (0.80) and integrity (0.73) violations and achieve 77% and 74% complexity reduction while maintaining 88% and 97% of the original performance of the secrecy and integrity violation prediction, respectively.

Paper Nr:	99
Title:	MK-SEAC: Multi-Keyword Searchable Encryption with Access Control
Authors:	Riccardo Longo, Enrico Sorbera and Valeria Vicard
Abstract:	In this work we present MK-SEAC, a scheme that allows to perform outsourced keyword-based search queries on an encrypted document batch, with native support to multi-keyword queries, fine-grained access control on the results, and full verifiability. The scheme conceals both the query and the results from the server, and minimizes the information revealed about documents to which users do not have access. Compared to other schemes, such as SEAC by Nils Löken, MK-SEAC offers extended improved efficiency with a reduced leakage profile.

Paper Nr:	113
Title:	A Fragile Watermarking Technique for Integrity Authentication of CSV-Files Using Invisible Line-Ending Control Characters
Authors:	Florian Zimmer, Malte Hellmeier, Motoki Nakamura and Tobias Urbanek
Abstract:	Every day, a growing amount of data, including audio, video, images, and plain text, is published and shared online. Facilitating its interoperable exchange, a range of standards and formats has emerged, establishing common ground. Among plain text formats, CSV prevails as one of the most used text formats. However, being a simplistic, plain text format, it lacks built-in security measures. Consequently, data users cannot authenticate the integrity of CSV texts they receive. A recognised method in research for ensuring text integrity is fragile watermarking. Accordingly, numerous watermarking techniques are available for tamper detection. However, many of these methods are either incompatible with the CSV format or visible to the human eye. To address these shortcomings, we propose a novel fragile watermarking technique for CSV files. Using invisible line-ending control characters, we are able to embed any byte-encodable information into a CSV cover text, making it truly imperceptible. We evaluated our technique by conducting three experiments to benchmark robustness, capacity and imperceptibility and comparing it with existing solutions. We found that our technique successfully achieves complete imperceptibility in all cases. However, a limited capacity and line-ending normalisation sensitivity must be considered when applying it.

Paper Nr:	125
Title:	A Hybrid Approach to Improve the Intrusion Detection Systems Using Generative Artificial Intelligence and Deep Reinforcement Learning
Authors:	Ines Ben Makhlouf, Ghassen Kilani, Fehmi Jaafar and Haïfa Nakouri
Abstract:	In recent years, Artificial Intelligence (AI)-based tools have gained widespread adoption as AI-powered prompts have become increasingly sophisticated. As a result, the rise of AI-integrated websites has created a growing demand for more sophisticated tools to protect devices and networks, especially in light of the emergence of AI-generated malware. Indeed, numerous studies anticipated the threats posed by this type of malware and proposed a variety of solutions to address this issue. In this context, most research introducing generative AI frameworks deals with image-based data, prompting the need to analyze tabular network data. We propose AAE-DRL, an Intrusion Detection System (IDS) that utilizes generative AI and deep reinforcement learning to replicate and predict intrusion behavior. We demonstrate the advantages and limitations of combining reconstruction and adversarial learning objectives with Deep Reinforcement Learning (DRL) in terms of intrusion detection, data generation, and minority sampling. Our approach achieved 89% accuracy, 90% precision, 91% recall, 90% F1-score on the augmented dataset with a 97% Area Under the Curve (AUC).

Paper Nr:	126
Title:	Enhancing Anonymity for Electric Vehicles in the ISO 15118 Plug-and-Charge
Authors:	Nethmi Hettiarachchi, Kalikinkar Mandal and Saqib Hakak
Abstract:	ISO 15118 is a standard protocol family that enables the plug-and-charge (PnC) functionality in the electric vehicle (EV) charging architecture. To initiate a charging session, an EV must first authenticate to the charging point (CP) by establishing a TLS connection using its X.509 certificate, followed by authorisation and billing at the end of charging. In this work, we first analyse the privacy of EVs with respect to the information exchanged during the ISO 15118 authentication, charging authorization and billing procedure. We discovered a significant privacy leakage in the current standard, where the initial authentication phase and the billing expose sensitive information that enables various attacks such as charging session linking, EV fingerprinting, profiling and resumption attacks against EV. To address this privacy issue, we first propose an efficient mutual authentication protocol for ISO 15118 PnC that protects the privacy of EVs, including identity and location, against the CP. We analyse the security of our protocol using the Tamarin formal verification tool. The protocol is implemented with various standardised cryptographic schemes and evaluated on different device platforms.

Paper Nr:	134
Title:	A Safety-Centric Analysis and Benchmarks of Modern Open-Source Homomorphic Encryption Libraries
Authors:	Nges Brian Njungle, Milan Stojkov and Michel A. Kinsy
Abstract:	Homomorphic Encryption (HE) is a rapidly evolving field in secure computation, offering very strong security guarantees in privacy-preserving data processing. A large number of commercial systems that prioritize privacy depend on open-source HE libraries to ensure secure and confidential computation. However, the security of these open-source libraries remains questionable, as they do not demonstrate strong security assurances, such as formal verification, in their development process. In this work, we investigate security vulnerabilities and the efficiency of the implementations of the four main HE schemes in the most commonly used open-source HE libraries. To analyze security, we employ the SafeRewrite open-source dynamic analysis tool, which uses symbolic execution techniques to validate code correctness. The study reveals several security vulnerabilities, errors, and warnings in all of the libraries. In terms of performance, we assess the latency and scalability of the fundamental HE operations in these libraries. The results indicate that the Cheon-Kim-Kim-Song (CKKS) scheme is the fastest HE scheme, whereas OpenFHE is, on average, the best-performing HE library. Overall, this research underscores the significance of using secure development approaches and frameworks in implementing HE algorithms to ensure stronger security guarantees and correctness while minimizing performance impacts.

Paper Nr:	137
Title:	MATRIX: A Comprehensive Graph-Based Framework for Malware Analysis and Threat Research
Authors:	Marco Simoni and Andrea Saracino
Abstract:	This paper presents MATRIX (Malware Analysis and Threat Research with STIX), a graph database for the comprehensive analysis and research of malware and threats. To provide a unified view of the threat landscape, MATRIX integrates data from major cybersecurity frameworks, including MITRE ATT&CK, DEF3ND, CAPEC, Malware Behavior Catalog (MBC), Metasploit, Common Vulnerabilities and Exposures (CVE) and Common Weakness Enumeration (CWE). Developed in Neo4j using the Structured Threat Information Expression (STIX™) standard, MATRIX includes more than 22,910 nodes and combines 14 STIX Domain Objects (SDOs) and 6 STIX Relationship Objects (SROs) to provide a detailed analysis of malware behavior, detection rules and defense strategies, making it a valuable tool for cybersecurity research. The system also integrates real-world malware reports and is automatically updated with data from sources such as VirusTotal, Malware-Bazaar and VirusShare, supporting continuous and up-to-date threat analysis. We demonstrate its versatility through case studies comparing malware objectives and analyzing the impact of detection and mitigation.

Paper Nr:	143
Title:	Enhancing Access Control in Distributed Systems Through Intelligent ABAC Policy Mining
Authors:	Sudhir Kumar Bai, Jason Aaron Goveas and Barsha Mitra
Abstract:	Distributed systems require secure, flexible, and efficient access control mechanisms to protect their resources and data. Attribute-Based Access Control (ABAC) has been found to be suitable for dynamic and cooperative settings of distributed environments. The successful implementation of ABAC in any system requires the formulation of a complete and correct ABAC policy. Creating a policy for ABAC adoption requires a substantial amount of computation and administrative effort. The scale of computational requirements and administrative efforts is further magnified if the target system of deployment is distributed in nature. Several heuristic methods have been proposed for ABAC policy generation. The amount of resources and efforts that need to be invested in policy formulation can be substantially reduced by leveraging machine learning techniques. In this paper, we propose an intelligent framework for mining ABAC policies from access logs for distributed systems. The task of policy generation is carried out in two phases. In the first phase, an initial policy is created by each of the individual entities of the distributed system. In the second phase, all the individually created policies are combined together to create the final ABAC policy. The proposed framework ensures data privacy by preventing the need for an entity to share its access log with any other entity by leveraging Federated Learning (FL) to create the ABAC policy. Experimental results on three access control datasets show that our proposed strategy creates ABAC policies which can efficiently and effectively evaluate access requests and perform access decision inferencing.

Paper Nr:	149
Title:	SONNI: Secure Oblivious Neural Network Inference
Authors:	Luke Sperling and Sandeep S. Kulkarni
Abstract:	In the standard privacy-preserving Machine learning as-a-service (MLaaS) model, the client encrypts data using homomorphic encryption and uploads it to a server for computation. The result is then sent back to the client for decryption. It has become more and more common for the computation to be outsourced to thirdparty servers. In this paper we identify a weakness in this protocol that enables a completely undetectable novel model-stealing attack that we call the Silver Platter attack. This attack works even under multikey encryption that prevents a simple collusion attack to steal model parameters. We also propose a mitigation that protects privacy even in the presence of a malicious server and malicious client or model provider (majority dishonest). When compared to a state-of-the-art but small encrypted model with 32k parameters, we preserve privacy with a failure chance of 1.51×10−28 while batching capability is reduced by 0.2%. Our approach uses a novel results-checking protocol that ensures the computation was performed correctly without violating honest clients’ data privacy. Even with collusion between the client and the server, they are unable to steal model parameters. Additionally, the model provider cannot learn any client data if maliciously working with the server.

Paper Nr:	152
Title:	Hierarchical Colored Petri Nets for Vulnerability Detection in Software Architectures
Authors:	Maya Benabdelhafid, Kamel Adi, Omer Landry Nguena Timo and Luigi Logrippo
Abstract:	Hierarchical Colored Petri Nets (HCPNs) are a powerful formalism for modeling complex systems. This paper presents a formal approach based on HCPN for vulnerability detection in software architecture. By incorporating model checking and the enhanced computing-timing logic of ASK-CTL queries, the proposed approach enables rigorous security property verification. Through a case study of a hypothetical small library system, we demonstrate how this automated process effectively identifies a critical Access Control vulnerability: a regular user gaining unauthorized access to a function reserved for librarians.

Paper Nr:	158
Title:	Optimal Noise Injection on Training Data: A Defense Against Membership Inference Attacks
Authors:	Radia Kassa, Kamel Adi and Myria Bouhaddi
Abstract:	Membership inference attacks (MIAs) present a serious risk to data privacy in machine learning (ML) models, as they allow attackers to determine whether a given data point was included in the training set. Although various defenses exist, they often struggle to effectively balance privacy and utility. To address this challenge, we propose in this paper a novel defense mechanism based on Optimal Noise Injection during the training phase. Our approach involves injecting a carefully designed and controlled noise vector into each training sample. This optimization maximizes prediction entropy to obscure membership signals while leveraging Shapley values to preserve data utility. Experiments on benchmark datasets show that our method reduces MIA success rates significantly without sacrificing accuracy, offering a strong privacy-utility trade-off for black-box scenarios.

Paper Nr:	159
Title:	Robust Peer-to-Peer Machine Learning Against Poisoning Attacks
Authors:	Myria Bouhaddi and Kamel Adi
Abstract:	Peer-to-Peer Machine Learning (P2P ML) offers a decentralized alternative to Federated Learning (FL), removing the need for a central server and enhancing scalability and privacy. However, the lack of centralized oversight exposes P2P ML to model poisoning attacks, where malicious peers inject corrupted updates. A major threat comes from adversarial coalitions, groups of peers that collaborate to reinforce poisoned updates and bypass local trust mechanisms. In this work, we investigate the impact of such coalitions and propose a defense framework that combines variance-based trust evaluation, Byzantine-inspired thresholding, and a feedback-driven self-healing mechanism. Extensive simulations in various attack scenarios demonstrate that our approach significantly improves robustness, ensuring high accuracy, detection by attackers, and model stability under adversarial conditions.

Paper Nr:	160
Title:	Adapting Under Fire: Multi-Agent Reinforcement Learning for Adversarial Drift in Network Security
Authors:	Emilia Rivas, Sabrina Saika, Ahtesham Bakht, Aritran Piplai, Nathaniel D. Bastian and Ankit Shah
Abstract:	Evolving attacks are a critical challenge for the long-term success of Network Intrusion Detection Systems (NIDS). The rise of these changing patterns has exposed the limitations of traditional network security methods. While signature-based methods are used to detect different types of attacks, they often fail to detect unknown attacks. Moreover, the system requires frequent updates with new signatures as the attackers are constantly changing their tactics. In this paper, we design an environment where two agents improve their policies over time. The adversarial agent, referred to as the red agent, perturbs packets to evade the intrusion detection mechanism, whereas the blue agent learns new defensive policies using drift adaptation techniques to counter the attacks. Both agents adapt iteratively: the red agent responds to the evolving NIDS, while the blue agent adjusts to emerging attack patterns. By studying the model’s learned policy, we offer concrete insights into drift adaptation techniques with high utility. Experiments show that the blue agent boosts model accuracy by 30% with just 2–3 adaptation steps using only 25–30 samples each.

Paper Nr:	11
Title:	A Method for Packed (and Unpacked) Malware Detection by Means of Convolutional Neural Networks
Authors:	Giovanni Ciaramella, Fabio Martinelli, Antonella Santone and Francesco Mercaldo
Abstract:	The current signature-based mechanism implemented by free and commercial antimalware requires the presence of the signature of the malicious sample to provide protection, i.e., to detect malicious behavior. This is why malware writers are developing techniques that can change the syntax of the code but leave the semantics unchanged, i.e., the malware business logic. Among these techniques is the so-called packed malware, i.e., malware with binary code modified by packers, software aimed to pack software, compress it, and package it with a stub. It is a program capable of decompressing and executing it in memory. In this way, malware detected by antimalware is not even detected in the packed version. In this paper, we propose a technique to detect packed malware by exploiting convolutional neural networks. In a nutshell, the proposed method performs static analysis, i.e., it does not require running the application to detect the malicious samples: we start from the application’s binary code exploited to generate an image that represents the input for a set of deep learning classifiers. The classifiers aim to discern an application under analysis between trusted or (packed) malicious. In the experimental analysis, we consider three different packers (i.e., mpress, BEP, and gzexe) to generate packed malware, thus demonstrating the ability of the proposed method to detect packed and unpacked malware with interesting performances.

Paper Nr:	20
Title:	Anomaly Detection in ZkSync Transactions with Unsupervised Machine Learning
Authors:	Kamil Kaczyński and Aleksander Wiącek
Abstract:	This work proposes an anomaly detection model that consists of two different machine learning algorithms, One Class Support Vector Machine and Isolation Forest. The chosen dataset is a publicly available ZkSync data dump. Although there are several articles on anomaly detection using machine learning in blockchains, this one is the first to focus on an Ethereum ZkSync rollup. There were two tasks set. One was to find suspicious transactions in a snippet of the dataset, and the second was to detect possible DDOS attacks, where one vector corresponds to one day of the life of the network. Evaluation of models was based on calculation of accuracy, synthetic accuracy, ROC AUC and PR AUC metrics. The models were fine-tuned on synthetically generated data. The proposed designs show reasonably good performance. The paper can be used as an inspiration to conduct more research on zero-knowledge rollups, as they may have slightly different user behavior than on Ethereum. In addition, the paper provides valuable insight into feature engineering and data processing, which can be useful to some researchers.

Paper Nr:	55
Title:	NULLDect: A Dynamic Adaptive Learning Framework for Robust NULL Pointer Dereference Detection
Authors:	Tasmin Karim, Md. Shazzad Hossain Shaon, Md. Fahim Sultan, Alfredo Cuzzocrea and Mst Shapna Akter
Abstract:	The identification of null pointer dereference vulnerabilities has implications for software security and reliability, as well as satisfying market needs for user data protection. This study introduces NULLDect, an adaptive learning-based approach that addresses this issue using the CWE-476 (NULL Pointer Dereference) dataset. Such detection becomes essential for averting software failures and unforeseen events that could compromise system stability and security. The proposed approach combines the uses of Long-Short-Term Memory (LSTM) networks, attention mechanisms, and adaptive learning with callback techniques to produce a phenomenal accuracy rate of 0.806 by extracting features utilizing the CodeT5 paradigm. Furthermore, the work incorporates and evaluates advanced computational models, including CodeT5, BERT, UniXcoder, and NLP-based GloVe embeddings, to discover the most successful strategy for null pointer detection across many evaluation metrics. This adaptability improves model accuracy, robustness, and longevity. NULLDect’s synergistic combination of approaches defines it as a comprehensive and effective solution for detecting and mitigating NULL pointer dereference problems.

Paper Nr:	58
Title:	A Mobile Payment Scheme Using Biometric Identification with Mutual Authentication
Authors:	Jack Sturgess and Ivan Martinovic
Abstract:	Cashless payment systems offer many benefits over cash, but also have some drawbacks. Fake terminals, skimming, wireless connectivity, and relay attacks are persistent problems. Attempts to overcome one problem often lead to another—for example, some systems use QR codes to avoid skimming and connexion issues, but QR codes can be stolen at distance and relayed. In this paper, we propose a novel mobile payment scheme based on biometric identification that provides mutual authentication to protect the user from rogue terminals. Our scheme imposes only minimal requirements on terminal hardware, does not depend on wireless connectivity between the user and the verifier during the authentication phase, and does not require the user to trust the terminal until it has authenticated itself to the user. We show that our scheme is resistant against phishing, replay, relay, and presentation attacks.

Paper Nr:	63
Title:	EDQKD: Enhanced-Dynamic Quantum Key Distributions with Improved Security and Key Rate
Authors:	Nikhil Kumar Parida, Sarath Babu, Neeraj Panwar and Virendra Singh
Abstract:	Widely adopted public key cryptography algorithms such as RSA and Elliptic Curve Cryptography (ECC) are susceptible to Shor’s algorithm, necessitating the development of quantum-secure cryptographic solutions. Quantum Key Distribution (QKD) has emerged as a potential approach for secure communication in the quantum era. However, existing QKD protocols suffer from inefficiencies in key exchange rates and vulnerabilities to attacks such as Photon Number Splitting (PNS) and Beam Splitter attacks. This paper proposes two dynamic QKD schemes that enhances security and efficiency by employing a dynamically changing control key. The unpredictability of these control keys ensures stronger randomness and resistance against adversarial attacks. The proposed scheme achieves a key exchange rate of 87.5%, significantly surpassing the 50% rate of the widely used BB84 protocol. These improvements demonstrate the potential of the proposed approach as a secure and efficient solution for quantum communication networks.

Paper Nr:	64
Title:	Post-Quantum Digital Signature Algorithms on IoT: Evaluating Performance on LoRa ESP32 Microcontroller
Authors:	Mads Villum Nielsen, Magnus Raagaard Kjeldsen, Togu Turnip and Birger Andersen
Abstract:	As quantum computing continues to develop, traditional cryptographic schemes face increasing threats from quantum attacks, driving the need for post-quantum cryptographic (PQC) algorithms. This study evaluates the feasibility of PQC algorithms with higher NIST security level parameters on a resource-constrained IoT device, the LoRa ESP32 microcontroller. We benchmarked the performance of CRYSTALS-Dilithium, Falcon, SPHINCS+ across multiple NIST levels, measuring latency, memory usage and discussing parameter sizes. Additionally, we examined the communication overhead introduced by transmitting the larger-than-usual digital signatures over a wireless network. Our findings reveal significant performance disparities between the tested algorithms, with Dilithium demonstrating the fastest execution and Falcon balancing speed and memory efficiency – even at higher NIST security levels. In contrast, SPHINCS+ proved impractically slow for IoT applications. This research investigates practical considerations and challenges of deploying PQC digital signature algorithms on IoT devices.

Paper Nr:	65
Title:	Stegoslayer: A Robust Browser-Integrated Approach for Thwarting Stegomalware
Authors:	Rushikesh Kawale, Sarath Babu and Virendra Singh
Abstract:	Over the years, various threat groups (APTs) have exploited innocuous-looking images as carriers for malware payloads, data exfiltration, and covert command and control communication by utilizing steganographic and polyglot techniques. Due to the widespread use of browsers as entry points to the internet, they have become the primary targets of online attacks. The attackers use the browser as an initial vector for carrying out steganographic-based attacks due to the browser’s ability to execute JavaScript. Attackers leverage this feature to extract and run hidden payloads from polyglot and steganographic media. To the best of our knowledge, no existing work prevents stegomalware attacks exploiting web-browser vulnerabilities, even though modern browsers remain susceptible to such attacks. Thus, to counter stegomalware attacks, we propose a steganographic attack prevention algorithm, Stegoslayer. Stegoslayer is an image-cleaning web extension and technique that ensures the image is free of malicious content while maintaining its quality. We performed functional tests against F5, Outguess and Openstego steganographic algorithm and stegosploit stegomalware. Further, we analyzed the performance of Stegoslayer against the state-of-the-art prevention method, Stegowiper. The results indicate that the output image of Stegoslayer has 20% better PSNR value than stegowiper.

Paper Nr:	77
Title:	Assessing Security RISC: Analyzing Flush+Fault Attack on RISC-V Using gem5 Simulator
Authors:	Mahreen Khan, Maria Mushtaq, Renaud Pacalet and Ludovic Apvrille
Abstract:	Microarchitectural side-channel attacks exploit vulnerabilities such as cache behavior to leak sensitive data. These attacks have been extensively studied on x86 architectures but they remain less explored on RISC-V systems. A recent paper (Gerlach et al., 2023) demonstrated existing and novel microarchitectural attacks on RISC-V hardware platforms (C906, U74, C910, C908). This hardware-based analysis, while realistic, lacks the flexibility and detailed behavioral insights needed to fully understand these attacks. Simulation environments like gem5 (Lowe-Power, 2024) provide fine-grained control and diverse metrics to overcome this limitation and observe the attack in detail. In this paper, gem5 is used to explore Flush+Fault (Gerlach et al., 2023) side-channel attack on RISC-V architecture which was originally tested on RISC-V hardware. Through gem5, we analyze detailed insights of attack such as cache patterns, and timing behaviors. Our results demonstrate the gem5’s potential for advancing the understanding of RISC-V microarchitectural vulnerabilities and eventually for developing effective countermeasures.

Paper Nr:	82
Title:	UOV-Based Verifiable Timed Signature Scheme
Authors:	Erkan Uslu and Oğuz Yayla
Abstract:	Verifiable Timed Signatures (VTS) are cryptographic primitives that enable the creation of a signature that can only be retrieved after a specific time delay, while also providing verifiable evidence of its existence. This framework is particularly useful in blockchain applications. Current VTS schemes rely on signature algorithms such as BLS, Schnorr, and ECDSA, which are vulnerable to quantum attacks due to the vulnerability of the discrete logarithm problem to Shor’s Algorithm. We introduce VT-UOV, a novel VTS scheme based on the Salt-Unbalanced Oil and Vinegar (Salt-UOV) Digital Signature Algorithm. As a multivariate polynomial-based cryptographic primitive, Salt-UOV provides strong security against both classical and quantum adversaries.

Paper Nr:	83
Title:	EdgeFuzz: A Middleware-Based Security Testing Tool for Vulnerability Discovery in Distributed Computing Applications
Authors:	Mishaal Ahmed and Muhammad Ajmal Naz
Abstract:	Distributed computing faces many security challenges due to the nature of the distribution of connecting nodes. Fuzz testing has become a popular automated tool for finding software system bugs and vulnerabilities in distributed environments. Distributed systems are characterized by various components spread across different network nodes. Such systems exhibit intrinsic complexities due to scalability, coordinated concurrency, and heterogeneity. Implementing fuzzing in such environments introduces additional challenges, such as managing communication and synchronization between distributed nodes to ensure that fuzzing tasks are executed promptly and coherently. In this paper, we proposed a novel EdgeFuzz Fuzzer to discover vulnerabilities in the server nodes of the distributed network through middleware-based fuzzing. EdgeFuzz is a black-box Fuzzer that modifies incoming client requests and sends them to the server while monitoring if a crash has occurred. We employed EdgeFuzz to test distributed networking tools and found server-side code vulnerabilities in selected applications.

Paper Nr:	88
Title:	Highlighting Vulnerabilities in a Genomics Biocybersecurity Lab Through Threat Modeling and Security Testing
Authors:	Jared Sheldon, Isabelle Brown-Cantrell, Patrick Pape and Thomas Morris
Abstract:	Biocybersecurity, a specialty field applying modern cybersecurity developments to the bioeconomy, is garnering progressively more attention as concerns increase over the protection of bioeconomic data generated each year. Genomic data is a key data type that falls under the bioeconomy umbrella and can be protected health information, intellectual property, or research data, depending on the use case. To increase understanding of cybersecurity for genomic lab environments, a biocybersecurity laboratory was set up and threat modeling was conducted on it using the STRIDE threat modeling methodology. Potential attack techniques were then mapped using the MITRE ATT&CK enterprise matrix and attack trees were generated to sequentially show the steps of these attacks. Going a step further, the initial steps of an attack tree were attempted against a DNA sequencer in the biocybersecurity lab. While the results of this testing did not yield an exploitable vulnerability that could be used to further test the attack tree techniques, lessons learned along the way can be taken into account by future research projects pursuing similar goals.

Paper Nr:	97
Title:	Threshold Structure-Preserving Signatures with Randomizable Key
Authors:	Ahmet Ramazan Ağırtaş, Emircan Çelik, Sermin Kocaman, Fatih Sulak and Oğuz Yayla
Abstract:	Digital signatures confirm message integrity and signer identity, but linking public keys to identities can cause privacy concerns in anonymized settings. Signatures with randomizable keys can break this link, preserving verifiability without revealing the signer. While effective for privacy, complex cryptographic systems need to be modular structured for efficient implementation. Threshold structure-preserving signatures enable modular, privacy-friendly protocols. This work combines randomizable keys with threshold structure-preserving signatures to create a valid, modular, and unlinkable foundation for privacy-preserving applications.

Paper Nr:	98
Title:	Privacy-Preserving EEG Data Generation: A Federated Split Learning Approach Using Privacy-Adaptive Autoencoders and Secure Aggregation with GFlowNet
Authors:	Shouvik Paul and Garima Bajwa
Abstract:	EEG-based Brain-Machine Interfaces (BMI) are novel interaction paradigms used extensively in assistive technologies and neurorehabilitation. However, these interfaces pose significant privacy risks as they rely on unique neural patterns for their operation, which unintentionally reveal sensitive cognitive information and biometric identifiers without consent. Unlike traditional data, EEG signals are challenging to anonymize due to their complex, high-dimensional, and noise-sensitive nature. We present a novel approach to privacy-preserving EEG data generation, combining Federated Split Learning (FSL) with hierarchical privacy-adaptive autoencoders, secure aggregation, and Generative Flow Networks (GFlowNet). The hierarchical architecture of the autoencoder enables multi-level feature extraction, effectively capturing both spatial and temporal de-pendencies in the EEG signals. Using Rényi Differential Privacy (RDP) and adaptive noise scaling, our model anonymizes sensitive brain signals during data generation. The FSL architecture allows client-side processing of raw EEG data, followed by server-side reconstruction and synthetic data generation using GFlowNet. Secure aggregation further enhances privacy, ensuring that individual data contributions are protected even during client and server communication. Evaluations of our approach under various privacy budgets demonstrate a balanced privacy-utility trade-off.

Paper Nr:	102
Title:	AI-Based Anomaly Detection and Classification of Traffic Using Netflow
Authors:	Gustavo Gonzalez Granadillo and Nesrine Kaaniche
Abstract:	Anomalies manifest differently in network statistics, making it difficult to develop generalized models for normal network behaviors and anomalies. This paper analyzes various Machine Learning (ML) and Deep Learning (DL) algorithms employing supervised techniques for both binary and multi-class classification of network traffic. Experiments have been conducted using a validated NetFlow-based dataset containing over 31 million incoming and outgoing network connections of an IT infrastructure. Preliminary results indicate that no single model effectively detects all cyber-attacks. However, selected models for binary and multi-class classification show promising results, achieving performance levels of up to 99.9% in the best of the cases.

Paper Nr:	104
Title:	Accelerating PEGASUS by Applying NTRU-Based GSW-Like Encryption
Authors:	Shusaku Uemura and Kazuhide Fukushima
Abstract:	Fully homomorphic encryption (FHE) enables to execute operations on ciphertexts without decryption. This leads to an expectation on FHE to be applied to analyses of confidential data. Each of FHE schemes proposed thus far has its own strengths such as an ability to handle real numbers or execute arbitrary functions. Scheme switching enables to switch one FHE ciphertext to another, and enables to utilize both strengths. However, as scheme switching is computationally expensive, it is sometimes more efficient to use one scheme with approximation. Xiang et al. (CRYPTO23) proposed an efficient blind rotation technique which is used in a scheme switching method named PEGASUS. They use the NTRU-based GSW-like encryption to accelerate blind rotation, which can be applied to schemes that use the GSW encryption. This paper investigates the effects of the application of the NTRU-based GSW-like encryption to PEGASUS. We found that applying the NTRU-based GSW-like encryption to PEGASUS theoretically reduces the key size required to evaluate a look-up table by 43% and the number of multiplications of integers by 96%. We also confirmed through experiments that using the NTRU-based GSW-like encryption in PEGASUS accelerates the evaluation of a look-up table by 1.55 times.

Paper Nr:	105
Title:	A Multi-Model Approach to Enhance Automatic Matching of Vulnerabilities to Attack Patterns
Authors:	Marine Sauze-Kadar and Thomas Loubier
Abstract:	Security knowledge databases represent key information in the process of vulnerability assessment and test automation of industrial products. The CVE and CAPEC databases respectively describe vulnerabilities and attack patterns. Linking a CVE entry to CAPEC can facilitate the generation of test plans, in the context of product test automation. Unfortunately, the great majority of CVE have no direct references to CAPEC. Several research works have focused on matching automatically CVE and CAPEC by computing text similarity on their descriptions, evaluating various models, in particular the term frequency inverse document frequency (TF-IDF) technique and transformer-based models such as SBERT. Depending on CVE description characteristics and evaluation criteria, these models are likely to perform differently by capturing different information types: vocabulary, preprocessing methods, context around words, etc. Hence, we propose a new classifier-based approach to select the most adapted similarity computation model from a given selection to match a CVE description with linked CAPEC descriptions. We evaluate this method on a recent set of CVE with CAPEC labels and show an improvement of matching accuracy compared to state-of-the-art methods leveraging a single model to compute text similarity. Our results also highlight the bias in the training and test set of CVE-CAPEC pairs.

Paper Nr:	106
Title:	Evasive IPv6 Covert Channels: Design, Machine Learning Detection, and Explainable AI Evaluation
Authors:	Viet Anh Phan and Jan Jerabek
Abstract:	Adopting a dual approach, this paper presents a framework that integrates two complementary components: CovertGen6, a novel tool for generating realistic IPv6 covert channel attack packets, and a framework of detection system based on multiple machine learning models. CovertGen6 outperforms existing tools by producing diverse, evasive attack scenarios that are captured by Wireshark and converted into CSV datasets for analysis. These authentic datasets are then used to train and evaluate machine learning models for detecting IPv6 covert channels, with the Random Forest classifier achieving a binary classification AuC of 0.985 and a multi-label classification F1-score of 90.3%. Additionally, the explainable AI technique is incorporated to transparently interpret model decisions and pinpoint the specific header fields used for covert injections. This dual approach bridges the gap between theoretical research and practical network security, laying a robust foundation for intrusion detection systems in IPv6 networks.

Paper Nr:	108
Title:	Dataset Watermarking Using the Discrete Wavelet Transform
Authors:	Mike P. Raave, Devriş İşler and Zekeriya Erkin
Abstract:	In this work, we focus on watermarking time series datasets and explore one of the techniques known from audio-watermarking, namely Discrete Wavelet Transform (DWT) based watermarking, to investigate its effectiveness. We adapt (Attari and A. Shirazi, 2018) and embed a bit stream into a time series dataset by calculating the DWT coefficients and modifying their magnitudes for embedding. Our experimental results on two real-world datasets show good robustness against a small range of data modification attacks but lack capability in larger-scale attacks. We believe that our work could initiate a new research direction on dataset watermarking using well-known techniques from signal processing.

Paper Nr:	111
Title:	Efficient Post-Processing of Intrusion Detection Alerts Using Data Mining and Clustering
Authors:	Kalu Gamage Kavindu Induwara Kumarasinghe, Ilangan Pakshage Madhawi Pathum Kumarsiri, Harsha Sandaruwan Gardiyawasam Pussewalage, Kapuruka Abarana Gedara Thihara Vilochana Kumarasinghe, Kushan Sudheera Kalupahana Liyanage, Yahani Pinsara Manawadu and Haran Mamankaran
Abstract:	Network Intrusion Detection Systems (IDS) have evolved significantly over the past two decades to address the growing complexity of network infrastructures and the increasing volume of cyber threats. However, traditional IDS approaches either rely on predefined signatures, which fail to detect zero-day attacks, or use anomaly detection models that suffer from high false alarm rates, overwhelming security analysts with excessive alerts. This paper proposes a data mining and adaptive clustering-based unsupervised approach to efficiently process IDS-generated network alerts, reducing false positives and enhancing threat detection. Relevant alert features are extracted, and advanced data mining techniques are applied to identify frequent patterns, reducing false alerts. Clustering similar patterns further groups alerts from related attacks, thereby reducing the workload of security analysts. This allows analysts to gain a high-level understanding of intrusions without manually reviewing vast numbers of alerts. The approach furthur enhances intrusion detection accuracy and provides actionable insights through alert correlation. The experimental results demonstrate significant improvements in detecting various cyber threats, including DDoS, Botnets, Port-scans, and more.

Paper Nr:	115
Title:	Software Benchmarking of NIST Lightweight Hash Function Finalists on Resource-Constrained AVR Platform via ChipWhisperer
Authors:	Mohsin Khan, Håvard Dagenborg and Dag Johansen
Abstract:	This paper presents a novel performance evaluation of five key lightweight hash functions on an ATxmega128 microcontroller, using our E-RANK metric as a composite metric that integrates execution speed, memory footprint, and energy efficiency into a unified and balanced ranking. We leverage hardware-specific profiling techniques, where counter registers are accessed directly on the microcontroller to measure execution speed and analyze RAM and ROM footprints post-compilation to determine memory usage. Energy consumption is measured using the ChipWhisperer FPGA toolkit, capturing voltage traces via an oscilloscope probe. Our evaluations provide new insights into the trade-offs inherent in each lightweight hash function, providing guidance on which one is most suitable for various application-specific constraints.

Paper Nr:	116
Title:	How to Design a Public Key Infrastructure for a Central Bank Digital Currency
Authors:	Makan Rafiee and Lars Hupel
Abstract:	Central Bank Digital Currency (CBDC) is a new form of money, issued by a country’s or region’s central bank, that can be used for a variety of payment scenarios. Depending on its concrete implementation, there are many participants in a production CBDC ecosystem, including the central bank, commercial banks, merchants, individuals, and wallet providers. There is a need for robust and scalable Public Key Infrastructure (PKI) for CBDC to ensure the continued trust of all entities in the system. This paper discusses the criteria that should flow into the design of a PKI and proposes a certificate hierarchy, together with a rollover concept ensuring continuous operation of the system. We further consider several peculiarities, such as the circulation of offline-capable hardware wallets.

Paper Nr:	119
Title:	An ETSI GS QKD Compliant TLS Implementation
Authors:	Thomas Prévost, Bruno Martin and Olivier Alibart
Abstract:	A modification of the TLS protocol is presented, using our implementation of the Quantum Key Distribution (QKD) standard ETSI GS QKD 014 v1.1.1. We rely on the Rustls library for this. The TLS protocol is modified while maintaining backward compatibility on the client and server side. We thus wish to participate in the effort to generalize the use of QKD on the Internet. We used our protocol for a video conference call encrypted by QKD. Finally, we analyze the performance of our protocol, comparing the time needed to establish a handshake to that of TLS 1.3.

Paper Nr:	122
Title:	From Real to Synthetic: GAN and DPGAN for Privacy Preserving Classifications
Authors:	Mohammad Emadi, Vahideh Moghtadaiee and Mina Alishahi
Abstract:	Generative Adversarial Networks (GANs) and Differentially Private GANs (DPGANs) have emerged as powerful tools for generating synthetic datasets while preserving privacy. In this work, we investigate the impact of using GAN- and DPGAN-generated datasets on the performance of machine learning classifiers. We generate synthetic datasets using both models and train a variety of classifiers to evaluate their accuracy and robustness on multiple benchmark datasets. We compare classifier performance on real versus synthetic datasets in four different evaluation scenarios. Our results provide insights into the feasibility of using GANs and DPGANs for privacy-preserving data generation and their implications for machine learning tasks.

Paper Nr:	123
Title:	New Integral Distinguishers and Security Reassessment of LTLBC
Authors:	Abhilash Kumar Das
Abstract:	This paper presents an in-depth study of integral distinguishers for the LTLBC block cipher, a 14-round 64-bit lightweight cryptographic scheme designed for low-latency applications in IoT environments. Leveraging the division property technique introduced by Yosuke Todo, we employ a Mixed Integer Linear Programming (MILP) approach to identify previously unpublished 6-round integral distinguishers for LTLBC. Additionally, we studied the MixWord permutation phase and showed that the cyclic intermixing tweak to the input word doesn’t yield any significant improvement. Instead, it reduces to the original MixWord operation. This observation is rigorously justified through an algebraic proof and further followed by linear and differential cryptanalysis, leading to a revised active Sbox count for LTLBC. As a side contribution, we correct inaccuracies in the reported division property propagations for the FUTURE Sbox, initially presented at AFRICACRYPT 2022. Our findings provide a deeper understanding of LTLBC’s security and offer valuable insights for the design of future lightweight block ciphers.

Paper Nr:	131
Title:	Enhancing National Digital Identity Systems: A Framework for Institutional and Technical Harm Prevention Inspired by Microsoft’s Harms Modeling
Authors:	Giovanni Corti, Gianluca Sassetti, Amir Sharif, Roberto Carbone and Silvio Ranise
Abstract:	The rapid adoption of National Digital Identity systems (NDIDs) across the globe underscores their role in ensuring the human right to identity. Despite the transformation potential given by digitization, these systems introduce significant challenges, particularly concerning their safety and potential misuse. When not adequately safeguarded, these technologies can expose individuals and populations to privacy risks as well as violations of their rights. These risks often originate from design and institutional flaws embedded in identity management infrastructures. Existing studies on NDIDs related harms often focus narrowly on technical design issues while neglecting the broader institutional infrastructures that enable such harms. To fill this gap, this paper extends the collection of harms for analysis through a qualitative methodology approach of the existing harm-related literature. Our findings suggest that 80% of NDID-related harms are the product of suboptimal institutions and poor governance models, and that 47.5% of all impacted stakeholders are considered High Risk. By proposing a more accurate harm assessment model, this paper provides academia and the industry with a significant contribution that allows for identifying the possibility of NDID-related harms at an embryonic state and building the necessary infrastructure to prevent them.

Paper Nr:	132
Title:	MorphDet: Towards the Detection of Morphing Attacks
Authors:	Jival Kapoor, Priyanka Singh and Manoranjan Mohanty
Abstract:	Biometric authentication systems have become an inevitable part of the society. They are based on the primary traits of an individual that are unique and hard to forge or manipulate by simple means. However, the unprecedented growth of technology has enabled the access of so many advanced tools that could be used for forging these traits. In this paper, we focus on the face morphing attacks. A basic pipeline is used to generate morphed attacks. A face morph detection model based on Resnet-152 is proposed and validated through exhaustive experiments. A dataset of 28,890 images is also contributed to conduct the experiments for varied scenarios, including simple face images, faces with beards, faces with eyeglasses, and a combination of beard and eyeglasses. Comparative performance analysis is done with the other state-of-the-art models i.e. Alexnet and VGG-16 and the proposed framework is found to outperform them.

Paper Nr:	138
Title:	Learning Without Sharing: A Comparative Study of Federated Learning Models for Healthcare
Authors:	Anja Campmans, Mina Alishahi and Vahideh Moghtadaiee
Abstract:	Federated Learning (FL) has emerged as a powerful approach for training machine learning (ML) models on decentralized healthcare data while maintaining patient privacy. However, selecting the most suitable FL model remains a challenge due to inherent trade-offs between accuracy and privacy. This study presents a comparative analysis of multiple FL optimization strategies applied to two real-world tabular health datasets. We evaluate the performance of FL models in terms of predictive accuracy, and resilience to privacy threats.Our findings provide insights into the practical deployment of FL in healthcare, highlighting key trade-offs and offering recommendations for selecting suitable FL models based on specific privacy and accuracy requirements.

Paper Nr:	140
Title:	Comparison of Credential Status Mechanisms for the Digital Wallet Ecosystem
Authors:	Riccardo Germenia, Salvatore Manfredi, Giada Sciarretta, Mario Scuro and Alessandro Tomasi
Abstract:	Digital identity wallets enable citizens to verify their identity and manage digital credentials. A system that offers the possibility of using and presenting credentials, requires the ability to check for their validity, avoiding the use of revoked or suspended credentials. This paper compares traditional and emerging credential status mechanisms to identify the most suitable solutions for the wallet ecosystem, taking in consideration privacy aspects and the set of available features.

Paper Nr:	146
Title:	Masked Vector Sampling for HQC
Authors:	Maxime Spyropoulos, David Vigilant, Fabrice Perion, Renaud Pacalet and Laurent Sauvage
Abstract:	Anticipating the advent of large quantum computers, NIST started a worldwide competition in 2016 aiming to define the next cryptographic standards. HQC is one of these post-quantum schemes selected for standardization. In 2022, Guo et al. introduced a timing attack that exploited a weakness in HQC rejection sampling function to recover its secret key in 866,000 calls to an oracle. The authors of HQC updated its specification by applying an algorithm to sample vectors in constant time. A masked implementation of this function was later proposed for BIKE but it is not directly applicable to HQC. In this paper we propose a specification-compliant masked version of the HQC vector sampling which relies, to our knowledge, on the first masked implementation of the Barrett reduction.

Paper Nr:	161
Title:	Seamless Post-Quantum Transition: Agile and Efficient Encryption for Data-at-Rest
Authors:	Federico Valbusa, Stephan Krenn, Thomas Lorünser and Sebastian Ramacher
Abstract:	As quantum computing advances, its threat to traditional cryptographic protocols, especially for long-term encrypted data, becomes critical. This paper presents an agile cryptosystem designed to ease the transition from pre-quantum to post-quantum security by supporting efficient integration of post-quantum Key Encapsulation Mechanisms (KEMs). Our approach combines a CCA-secure KEM with robust Authenticated Encryption (AE), allowing only the encapsulated key to be updated during migration, without re-encrypting large data payloads—saving both computation and bandwidth. We formalize cryptographic agility via an agile-CCA security model, ensuring that neither the original nor updated ciphertexts leak information. A game-based proof shows that the construction remains agile-CCA secure if the underlying KEM and AE are individually CCA-secure in the random oracle model. The result is a future-proof scheme that enables enterprises and cloud providers to safeguard vast data volumes against emerging quantum threats with minimal disruption.

Paper Nr:	162
Title:	Privacy-Enhancing Federated Time-Series Forecasting: A Microaggregation-Based Approach
Authors:	Sargam Gupta and Vicenç Torra
Abstract:	Time-series forecasting is predicting future values based on historical data. Applications include forecasting traffic flows, stock market trends, and energy consumption, which significantly helps to reduce costs and efficiency. However, the complexity inherent in time-series data makes accurate forecasting challenging. This article proposes a novel privacy-enhancing k-anonymous federated learning framework for time-series prediction based on microaggregation. This adaptable framework can be customised based on the client-side processing capabilities. We evaluate the performance of our proposed framework by comparing it with the centralized one using the standard metrics like Mean Absolute Error on three real-world datasets. Moreover, we performed a detailed ablation study by experimenting with different values of k in microaggregation and different client side forecasting models. The results show that our approach gives comparable a good privacy-utility tradeoff as compared to the centralized benchmark.

Paper Nr:	163
Title:	Extending Null Embedding for Deep Neural Network (DNN) Watermarking
Authors:	Kaan Altınay, Devriş İşler and Zekeriya Erkin
Abstract:	The rise of Machine Learning (ML) has opened new business opportunities, particularly through Machine Learning as a Service (MLaaS), where costly models like Deep Neural Networks (DNNs) can be outsourced. However, this also raises concerns about model piracy. To protect against unauthorized use, watermarking techniques have been developed. One such method, null embedding by Li et al., disables the model if pirated but reduces classification accuracy. This paper proposes modifications to the null-embedding technique that reduce this impact and keep the classification accuracy close to that of a non-watermarked model.

Paper Nr:	170
Title:	Supporting Resilient, Ethical, and Verifiable Anonymous Identities Through Blockchains
Authors:	Alberto De Marchi, Lorenzo Gigli, Andrea Melis, Luca Sciullo and Fabio Vitali
Abstract:	In recent years, anonymity on the internet has come under intense scrutiny for enabling criminal behaviors like cyberbullying, disinformation, child exploitation, and illicit financial activities. Nevertheless, strong advocates highlight its importance as a protective space for legitimate and ethical actions that individuals may prefer to keep separate from their real-world identities. This paper presents a protocol for authenticated anonymity, enabling anonymous usage that remains unlinkable to real identities unless criminal activity is detected. Blockchain offers a robust and secure framework to manage these needs. While existing solutions — e.g., self-sovereign identities — grant users full control over their disclosure, they lack proper accountability. To address this limitation, the proposed protocol employs a blockchain-driven mechanism that supports anonymous yet verifiable identities. De-anonymization is achieved exclusively through multi-party consensus on the blockchain, triggered by explicit and non-repudiable requests. We provide the formal mathematical model of the protocol and offer some evaluations of its robustness and fault tolerance, even under large-scale identity management scenarios.