Papers

Deceptive UI designs, widely instantiated across the web and commonly known as dark patterns, manipulate users into performing actions misaligned with their goals. In this paper, we show that dark patterns are highly effective in steering agent trajectories, posing a significant risk to agent robustness. To quantify this risk, we introduce DECEPTICON, an environment for testing individual dark patterns in isolation. DECEPTICON includes 700 web navigation tasks with dark patterns – 600 generated tasks and 100 real-world tasks, designed to measure instruction-following success and dark pattern effectiveness. Across state-of-the-art agents, we find dark patterns successfully steer agent trajectories towards malicious outcomes in over 70% of tested generated and real-world tasks – compared to a human average of 31%. Moreover, we find that dark pattern effectiveness correlates positively with model size and test-time reasoning, making larger, more capable models more susceptible. Leading countermeasures against adversarial attacks, including in-context prompting and guardrail models, fail to consistently reduce the success rate of dark pattern interventions. Our findings reveal dark patterns as a latent and unmitigated risk to web agents, highlighting the urgent need for robust defenses against manipulative designs.

Cybersecurity is being fundamentally reshaped by foundation-model-based artificial intelligence. Large language models now enable autonomous planning, tool orchestration, and strategic adaptation at scale, challenging security architectures built on static rules, perimeter defenses, and human-centered workflows. This chapter argues for a shift from prevention-centric security toward agentic cyber resilience. Rather than seeking perfect protection, resilient systems must anticipate disruption, maintain critical functions under attack, recover efficiently, and learn continuously. We situate this shift within the historical evolution of cybersecurity paradigms, culminating in an AI-augmented paradigm where autonomous agents participate directly in sensing, reasoning, action, and adaptation across cyber and cyber-physical systems. We then develop a system-level framework for designing agentic AI workflows. A general agentic architecture is introduced, and attacker and defender workflows are analyzed as coupled adaptive processes, and game-theoretic formulations are shown to provide a unifying design language for autonomy allocation, information flow, and temporal composition. Case studies in automated penetration testing, remediation, and cyber deception illustrate how equilibrium-based design enables system-level resiliency design.

Securing blockchain-enabled IoT networks against sophisticated adversarial attacks remains a critical challenge. This paper presents a trust-based delegated consensus framework integrating Fully Homomorphic Encryption (FHE) with Attribute-Based Access Control (ABAC) for privacy-preserving policy evaluation, combined with learning-based defense mechanisms. We systematically compare three reinforcement learning approaches – tabular Q-learning (RL), Deep RL with Dueling Double DQN (DRL), and Multi-Agent RL (MARL) – against five distinct attack families: Naive Malicious Attack (NMA), Collusive Rumor Attack (CRA), Adaptive Adversarial Attack (AAA), Byzantine Fault Injection (BFI), and Time-Delayed Poisoning (TDP). Experimental results on a 16-node simulated IoT network reveal significant performance variations: MARL achieves superior detection under collusive attacks (F1=0.85 vs. DRL's 0.68 and RL's 0.50), while DRL and MARL both attain perfect detection (F1=1.00) against adaptive attacks where RL fails (F1=0.50). All agents successfully defend against Byzantine attacks (F1=1.00). Most critically, the Time-Delayed Poisoning attack proves catastrophic for all agents, with F1 scores dropping to 0.11-0.16 after sleeper activation, demonstrating the severe threat posed by trust-building adversaries. Our findings indicate that coordinated multi-agent learning provides measurable advantages for defending against sophisticated trust manipulation attacks in blockchain IoT environments.

Recent privacy regulations such as the General Data Protection Regulation (GDPR) and the California Consumer Privacy Act (CCPA) have established legal requirements for obtaining user consent regarding the collection, use, and sharing of personal data. These regulations emphasize that consent must be informed, freely given, specific, and unambiguous. However, there are still many violations, which highlight a gap between legal expectations and actual implementation. Consent mechanisms embedded in functional web forms across websites play a critical role in ensuring compliance with data protection regulations such as the GDPR and CCPA, as well as in upholding user autonomy and trust. However, current research has primarily focused on cookie banners and mobile app dialogs. These forms are diverse in structure, vary in legal basis, and are often difficult to locate or evaluate, creating a significant challenge for automated consent compliance auditing. In this work, we present Cosmic, a novel automated framework for detecting consent-related privacy violations in web forms. We evaluate our developed tool for auditing consent compliance in web forms, across 5,823 websites and 3,598 forms. Cosmic detects 3,384 violations on 94.1% of consent forms, covering key GDPR principles such as freely given consent, purpose disclosure, and withdrawal options. It achieves 98.6% and 99.1% TPR for consent and violation detection, respectively, demonstrating high accuracy and real-world applicability.

Signed networks, encoding both positive and negative interactions, are essential for modeling complex systems in social and financial domains. Sign prediction, which infers the sign of a target link, has wide-ranging practical applications. Traditional motif-based Naïve Bayes models assume that all neighboring nodes contribute equally to a target link's sign, overlooking the heterogeneous influence among neighbors and potentially limiting performance. To address this, we propose a generalizable sign prediction framework that explicitly models the heterogeneity. Specifically, we design two role functions to quantify the differentiated influence of neighboring nodes. We further extend this approach from a single motif to multiple motifs via two strategies. The generalized multiple motifs-based Naïve Bayes model linearly combines information from diverse motifs, while the Feature-driven Generalized Motif-based Naïve Bayes (FGMNB) model integrates high-dimensional motif features using machine learning. Extensive experiments on four real-world signed networks show that FGMNB consistently outperforms five state-of-the-art embedding-based baselines on three of these networks. Moreover, we observe that the most predictive motif structures differ across datasets, highlighting the importance of local structural patterns and offering valuable insights for motif-based feature engineering. Our framework provides an effective and theoretically grounded solution to sign prediction, with practical implications for enhancing trust and security in online platforms.

Identifying social bots has become a critical challenge due to their significant influence on social media ecosystems. Despite advancements in detection methods, most topology-based approaches insufficiently account for the heterogeneity of neighborhood preferences and lack a systematic theoretical foundation, relying instead on intuition and experience. Here, we propose a theoretical framework for detecting social bots utilizing heterogeneous motifs based on the Naïve Bayes model. Specifically, we refine homogeneous motifs into heterogeneous ones by incorporating node-label information, effectively capturing the heterogeneity of neighborhood preferences. Additionally, we systematically evaluate the contribution of different node pairs within heterogeneous motifs to the likelihood of a node being identified as a social bot. Furthermore, we mathematically quantify the maximum capability of each heterogeneous motif, enabling the estimation of its potential benefits. Comprehensive evaluations on four large, publicly available benchmarks confirm that our method surpasses state-of-the-art techniques, achieving superior performance across five evaluation metrics. Moreover, our results reveal that selecting motifs with the highest capability achieves detection performance comparable to using all heterogeneous motifs. Overall, our framework offers an effective and theoretically grounded solution for social bot detection, significantly enhancing cybersecurity measures in social networks.

This paper explores vulnerabilities in RSA cryptosystems that arise from improper prime number selection during key generation. We examine two primary attack vectors: Fermat's factorization method, which exploits RSA keys generated with primes that are too close together, and the Greatest Common Divisor (GCD) attack, which exploits keys that share a common prime factor. Drawing from landmark research including Heninger et al.'s “Mining Your Ps and Qs” study, which discovered over 64,000 vulnerable TLS hosts, and Böck's 2023 analysis of Fermat factorization in deployed systems, we demonstrate that these vulnerabilities remain prevalent in real-world cryptographic implementations. Our analysis reveals that weak random number generation in embedded devices is the primary cause of these failures, and we discuss mitigation strategies including proper entropy collection and prime validation checks.

Understanding the risks associated with an enterprise environment is the first step toward improving its security. Organizations employ various methods to assess and prioritize the risks identified in cyber threat intelligence (CTI) reports that may be relevant to their operations. Some methodologies rely heavily on manual analysis (which requires expertise and cannot be applied frequently), while others automate the assessment, using attack graphs (AGs) or threat emulators. Such emulators can be employed in conjunction with cyber twins to avoid disruptions in live production environments when evaluating the highlighted threats. Unfortunately, the use of cyber twins in organizational networks is limited due to their inability to scale. In this paper, we propose SCyTAG, a multi-step framework that generates the minimal viable cyber twin required to assess the impact of a given attack scenario. Given the organizational computer network specifications and an attack scenario extracted from a CTI report, SCyTAG generates an AG. Then, based on the AG, it automatically constructs a cyber twin comprising the network components necessary to emulate the attack scenario and assess the relevance and risks of the attack to the organization. We evaluate SCyTAG on both a real and fictitious organizational network. The results show that compared to the full topology, SCyTAG reduces the number of network components needed for emulation by up to 85% and halves the amount of required resources while preserving the fidelity of the emulated attack. SCyTAG serves as a cost-effective, scalable, and highly adaptable threat assessment solution, improving organizational cyber defense by bridging the gap between abstract CTI and practical scenario-driven testing.

We frame Ethereum transactions reverted by invariants-require(<invariant>)/ assert(<invariant>)/if (<invariant>) revert statements in the contract implementation-as a positive signal of active on-chain defenses. Despite their value, the defensive patterns in these transactions remain undiscovered and underutilized in security research. We present Raven, a framework that aligns reverted transactions to the invariant causing the reversion in the smart contract source code, embeds these invariants using our BERT-based fine-tuned model, and clusters them by semantic intent to mine defensive invariant categories on Ethereum. Evaluated on a sample of 20,000 reverted transactions, Raven achieves cohesive and meaningful clusters of transaction-reverting invariants. Manual expert review of the mined 19 semantic clusters uncovers six new invariant categories absent from existing invariant catalogs, including feature toggles, replay prevention, proof/signature verification, counters, caller-provided slippage thresholds, and allow/ban/bot lists. To demonstrate the practical utility of this invariant catalog mining pipeline, we conduct a case study using one of the newly discovered invariant categories as a fuzzing oracle to detect vulnerabilities in a real-world attack. Raven thus can map Ethereum's successful defenses. These invariant categories enable security researchers to develop analysis tools based on data-driven security oracles extracted from the smart contracts' working defenses.

Modern cloud-based AI training relies on extensive telemetry and logs to ensure accountability. While these audit trails enable retrospective inspection, they struggle to address the inherent non-determinism of deep learning. Stochastic operations, such as dropout, create an ambiguity surface where attackers can mask malicious manipulations as natural random variance, granting them plausible deniability. Consequently, existing logging mechanisms cannot verify whether stochastic values were generated and applied honestly without exposing sensitive training data. To close this integrity gap, we introduce Verifiable Dropout, a privacy-preserving mechanism based on zero-knowledge proofs. We treat stochasticity not as an excuse but as a verifiable claim. Our approach binds dropout masks to a deterministic, cryptographically verifiable seed and proves the correct execution of the dropout operation. This design enables users to audit the integrity of stochastic training steps post-hoc, ensuring that randomness was neither biased nor cherry-picked, while strictly preserving the confidentiality of the model and data.

Ensuring the authenticity and ownership of digital images is increasingly challenging as modern editing tools enable highly realistic forgeries. Existing image protection systems mainly rely on digital watermarking, which is susceptible to sophisticated digital attacks. To address this limitation, we propose a hybrid optical-digital framework that incorporates physical authentication cues during image formation and preserves them through a learned reconstruction process. At the optical level, a phase mask in the camera aperture produces a Null-space Optical Watermark (NOWA) that lies in the Null Space of the imaging operator and therefore remains invisible in the captured image. Then, a Null-Space Network (NSN) performs measurement-consistent reconstruction that delivers high-quality protected images while preserving the NOWA signature. The proposed design enables tamper localization by projecting the image onto the camera's null space and detecting pixel-level inconsistencies. Our design preserves perceptual quality, resists common degradations such as compression, and establishes a structural security asymmetry: without access to the optical or NSN parameters, adversaries cannot forge the NOWA signature. Experiments with simulations and a prototype camera demonstrate competitive performance in terms of image quality preservation, and tamper localization accuracy compared to state-of-the-art digital watermarking and learning-based authentication methods.

We introduce LLA, an effective intellectual property (IP) protection scheme for generative AI models. LLA leverages the synergy between hardware and software to defend against various supply chain threats, including model theft, model corruption, and information leakage. On the software side, it embeds key bits into neurons that can trigger outliers to degrade performance and applies invariance transformations to obscure the key values. On the hardware side, it integrates a lightweight locking module into the AI accelerator while maintaining compatibility with various dataflow patterns and toolchains. An accelerator with a pre-stored secret key acts as a license to access the model services provided by the IP owner. The evaluation results show that LLA can withstand a broad range of oracle-guided key optimization attacks, while incurring a minimal computational overhead of less than 0.1% for 7,168 key bits.

Large Language Models (LLMs) have demonstrated significant potential in automated software security, particularly in vulnerability detection. However, existing benchmarks primarily focus on isolated, single-vulnerability samples or function-level classification, failing to reflect the complexity of real-world software where multiple interacting vulnerabilities often coexist within large files. Recent studies indicate that LLMs suffer from "count bias" and "selection bias" in multi-label tasks, yet this has not been rigorously quantified in the domain of code security. In this work, we introduce a comprehensive benchmark for Multi-Vulnerability Detection across four major languages: C, C++, Python, and JavaScript. We construct a dataset of 40,000 files by systematically injecting controlled counts of vulnerabilities (1, 3, 5, and 9) into long-context code samples (7.5k-10k tokens) sourced from CodeParrot. We evaluate five state-of-the-art LLMs, including GPT-4o-mini, Llama-3.3-70B, and the Qwen-2.5 series. Our results reveal a sharp degradation in performance as vulnerability density increases. While Llama-3.3-70B achieves near-perfect F1 scores (approximately 0.97) on single-vulnerability C tasks, performance drops by up to 40% in high-density settings. Notably, Python and JavaScript show distinct failure modes compared to C/C++, with models exhibiting severe "under-counting" (Recall dropping to less than 0.30) in complex Python files.

Timing side-channels are an important threat to cryptography that still needs to be addressed in implementations, and the advent of post-quantum cryptography raises this issue because the lattice-based schemes may produce secret-dependent timing variability with the help of complex arithmetic and control flow. Since also real timing measurements are affected by environmental noise (e.g. scheduling effects, contention, heavy tailed delays), in this work a scenario-based statistical risk model is proposed for timing leakage as a problem of distributional distinguishability under controlled execution conditions. We synthesize traces for two secret classes in idle, jitter and loaded scenarios and for multiple leakage models and quantify leakage with Welch's t-test, KS distance, Cliff's delta, mutual information, and distribution overlap to combine in a TLRI like manner to obtain a consistent score for ranking scenarios. Across representative lattice-based KEM families (Kyber, Saber, Frodo), idle conditions generally have the best distinguishability, jitter and loaded conditions erode distinguishability through an increase in variance and increase in overlap; cache-index and branch-style leakage tends to give the highest risk signals, and faster schemes can have a higher peak risk given similar leakage assumptions, allowing reproducible comparisons at an early design stage, prior to platform-specific validation.

Trusted Execution Environments (TEEs) protect sensitive code and data from the operating system, hypervisor, or other untrusted software. Different solutions exist, each proposing different features. Abstraction layers aim to unify the ecosystem, allowing application developers and system administrators to leverage confidential computing as broadly and efficiently as possible. We start with an overview of representative available TEE technologies. We describe and summarize each TEE ecosystem, classifying them in different categories depending on their main design choices. Then, we propose a systematization of knowledge focusing on different abstraction layers around each design choice. We describe the underlying technologies of each design, as well as the inner workings and features of each abstraction layer. Our study reveals opportunities for improving existing abstraction layer solutions. It also highlights WebAssembly, a promising approach that supports the largest set of features. We close with a discussion on future directions for research, such as how future abstraction layers may evolve and integrate with the confidential computing ecosystem.

Software obfuscation techniques make code more difficult to understand, without changing its functionality. Such techniques are often used by authors of malicious software to avoid detection. Reverse Engineering of obfuscated code, i.e., the process of overcoming obfuscation and answering questions about the functionality of the code, is notoriously difficult; and while various tools and methods exist for this purpose, the process remains complex and slow, especially when dealing with layered or customized obfuscation techniques. Here, we present a novel, automated tool for addressing some of the challenges in reverse engineering of obfuscated code. Our tool, called ReSMT, converts the obfuscated assembly code into a complex system of logical assertions that represent the code functionality, and then applies SMT solving and simulation tools to inspect the obfuscated code's execution. The approach is mostly automatic, alleviating the need for highly specialized deobfuscation skills. In an elaborate case study that we conducted, ReSMT successfully tackled complex obfuscated code, and was able to solve reverse-engineering queries about it. We believe that these results showcase the potential and usefulness of our proposed approach.

Natural Language Processing (NLP) systems are increasingly used in sensitive domains such as healthcare, finance, and government, where they handle large volumes of personal and regulated data. However, these systems introduce distinct risks related to security, privacy, and regulatory compliance that are not fully addressed by existing AI governance frameworks. This paper introduces the Secure and Compliant NLP Lifecycle Management Framework (SC-NLP-LMF), a comprehensive six-phase model designed to ensure the secure operation of NLP systems from development to retirement. The framework, developed through a systematic PRISMA-based review of 45 peer-reviewed and regulatory sources, aligns with leading standards, including NIST AI RMF, ISO/IEC 42001:2023, the EU AI Act, and MITRE ATLAS. It integrates established methods for bias detection, privacy protection (differential privacy, federated learning), secure deployment, explainability, and secure model decommissioning. A healthcare case study illustrates how SC-NLP-LMF detects emerging terminology drift (e.g., COVID-related language) and guides compliant model updates. The framework offers organizations a practical, lifecycle-wide structure for developing, deploying, and maintaining secure and accountable NLP systems in high-risk environments.

Prompt-driven Video Segmentation Foundation Models (VSFMs), such as SAM2, are increasingly used in applications including autonomous driving and digital pathology, yet their security risks remain underexplored. We study backdoor attacks against VSFMs and show that directly applying classic attacks such as BadNet is largely ineffective, yielding attack success rates (ASR) below 5%. Through gradient-similarity and attention-map analyses, we find that traditional backdoor training fails because clean and triggered samples induce aligned image-encoder gradients, while model attention remains focused on the prompt-specified object rather than the trigger. To address this limitation, we propose BadVSFM, the first backdoor attack framework tailored to prompt-driven VSFMs. BadVSFM uses a two-stage strategy that first learns trigger-specific encoder features and then trains the decoder to map triggered frame prompt representations to an attacker-specified target mask while preserving clean segmentation behavior. Experiments on five VSFMs and two datasets show that BadVSFM achieves strong, controllable backdoor effects across triggers and prompt types with limited clean-performance degradation. Ablations and interpretability analyses validate the necessity of the two-stage design, and five representative defenses remain largely ineffective. Our results reveal a practical and underexplored vulnerability of current VSFMs to backdoor threats.

Large vision-language models (LVLMs) have achieved remarkable advancements in multimodal reasoning tasks. However, their widespread accessibility raises critical concerns about potential copyright infringement. Will LVLMs accurately recognize and comply with copyright regulations when encountering copyrighted content (i.e., user input, retrieved documents) in the context? Failure to comply with copyright regulations may lead to serious legal and ethical consequences, particularly when LVLMs generate responses based on copyrighted materials (e.g., retrieved book experts, news reports). In this paper, we present a comprehensive evaluation of various LVLMs, examining how they handle copyrighted content – such as book excerpts, news articles, music lyrics, and code documentation when they are presented as visual inputs. To systematically measure copyright compliance, we introduce a large-scale benchmark dataset comprising 50,000 multimodal query-content pairs designed to evaluate how effectively LVLMs handle queries that could lead to copyright infringement. Given that real-world copyrighted content may or may not include a copyright notice, the dataset includes query-content pairs in two distinct scenarios: with and without a copyright notice. For the former, we extensively cover four types of copyright notices to account for different cases. Our evaluation reveals that even state-of-the-art closed-source LVLMs exhibit significant deficiencies in recognizing and respecting the copyrighted content, even when presented with the copyright notice. To solve this limitation, we introduce a novel tool-augmented defense framework for copyright compliance, which reduces infringement risks in all scenarios. Our findings underscore the importance of developing copyright-aware LVLMs to ensure the responsible and lawful use of copyrighted content.

The Internet of Drones (IoD) is an emerging and crucial paradigm enabling advanced applications that require seamless, secure communication across heterogeneous and untrusted domains. In such environments, access control and the transmission of sensitive data pose significant security challenges for IoD systems, necessitating the design of lightweight mutual authentication and key exchange protocols. Existing solutions are often hampered by high computation overhead, reliance on third parties, the requirement for secret storage in resource-constrained drones, and the need for a strictly controlled enrollment environment. These limitations make them impractical for dynamic cross-domain deployment. To address these limitations, we propose a lightweight mutual authentication mechanism that integrates Radio Frequency Fingerprint (RFF) and Physical Unclonable Function (PUF) technologies for secure drone-to-drone (D2D) and drone-to-ground station server (D2G) communication. RFF-based device identification is used to achieve over-the-air (OTA) enrollment, while the PUF serves as the root of trust for establishing mutual authentication among communication parties. Additionally, the on-the-fly key generation capability of the PUF is co-designed with One-Time-Pad (OTP) encryption to realize ephemeral keying and eliminate the need for storing secrets within drones. Both informal security analysis and ProVerif-based formal security verification comprehensively demonstrate the resilience of our protocol against common security attacks. The proposed protocol also outperforms existing IoD authentication schemes in terms of security features, as well as computation, communication, and storage overhead.

The Advanced Dynamic Security Learning (DSL) Process Model is an Industry 4.0 cybersecurity incident response architecture proposed in this paper. This model addresses proactive and reflective cybersecurity governance across complex cyber-physical systems by combining Argyris and Schön's double-loop learning theory with Crossan's 4I organizational learning framework. Given that 65% of industrial companies suffer ransomware attacks annually and many of them lack cybersecurity awareness, this reveals the gravity of cyber threats. Feedforward and feedback learning loops in this paradigm help promote strategic transformation and ongoing growth. The DSL model helps Industry 4.0 organizations adapt to growing challenges posed by the projected 18.8 billion IoT devices by bridging operational obstacles and promoting systemic resilience. This research presents a scalable, methodical cybersecurity maturity approach based on a comprehensive analysis of the literature and a qualitative study.

Large language models (LLMs) are highly compute- and memory-intensive, posing significant demands on high-performance GPUs. At the same time, advances in GPU technology driven by shrinking transistor sizes and lower operating voltages have made these devices increasingly susceptible to soft errors. While prior work has examined GPU reliability, most studies have focused on general-purpose applications or conventional neural networks mostly used for vision tasks such as classification and detection. In contrast, systematic analysis of modern large-scale LLMs remains limited, despite their rapid adoption in diverse application scenarios. Given the unique characteristics of LLMs, their resilience to soft errors may differ substantially from earlier models. To bridge this gap, we conduct the first instruction-level fault injection study of LLM inference. Our approach reveals reliability characteristics from multiple perspectives, highlighting the effects of model architecture, parameter scale, and task complexity. These findings provide new insights into LLM reliability and inform the design of more effective fault tolerance mechanisms.

Warning-framed content in training data (e.g., "DO NOT USE - this code is vulnerable") does not, it turns out, teach language models to avoid the warned-against behavior. In experiments reported here, models exposed to such warnings reproduced the flagged content at rates statistically indistinguishable from models given the content directly (76.7% vs. 83.3%). Why? Sparse autoencoder analysis points to a failure of orthogonalization: "describing X" and "performing X" activate overlapping latent features. Feature #8684, which tracks code execution patterns, fires at comparable magnitude in both warning and exploitation contexts. A related phenomenon, what I call "stealth slip", allows conversational preambles to rotate activations into subspaces that linear probes miss entirely. Prompting and inference-time steering do not fix this; training-time feature ablation does. The upshot is that statistical co-occurrence dominates over pragmatic interpretation in current architectures. Models learn what tends to follow a context, not why it appeared there.

Generative AI systems are quickly improving, now able to produce believable output in several modalities including images, text, and audio. However, this fast development has prompted increased scrutiny concerning user privacy and the use of copyrighted works in training. A recent attack on machine-learning models called membership inference lies at the crossroads of these two concerns. The attack is given as input a set of records and a trained model and seeks to identify which of those records may have been used to train the model. On one hand, this attack can be used to identify user data used to train a model, which may violate their privacy especially in sensitive applications such as models trained on medical data. On the other hand, this attack can be used by rights-holders as evidence that a company used their works without permission to train a model. Remarkably, it appears that no work has studied the effect of membership inference attacks (MIA) on generative music. Given that the music industry is worth billions of dollars and artists would stand to gain from being able to determine if their works were being used without permission, we believe this is a pressing issue to study. As such, in this work we begin a preliminary study into whether MIAs are effective on generative music. We study the effect of several existing attacks on MuseGAN, a popular and influential generative music model. Similar to prior work on generative audio MIAs, our findings suggest that music data is fairly resilient to known membership inference techniques.

This paper demonstrates a power analysis-based Side-Channel Analysis (SCA) attack on the SNOW-V encryption algorithm, which is a 5G mobile communication security standard candidate. Implemented on an STM32 microcontroller, power traces captured with a ChipWhisperer board were analyzed, with Test Vector Leakage Assessment (TVLA) confirming exploitable leakage. Profiling attacks using Linear Discriminant Analysis (LDA) and Fully Connected Neural Networks (FCN) achieved efficient key recovery, with FCN achieving > 5X lower minimum traces to disclosure (MTD) compared to the state-of-the-art Correlational Power Analysis (CPA) assisted with LDA. The results highlight the vulnerability of SNOW-V to machine learning-based SCA and the need for robust countermeasures.

This work introduces a unified raster domain steganographic framework, termed as the Glyph Perturbation Cardinality (GPC) framework, capable of embedding heterogeneous data such as text, images, audio, and video directly into the pixel space of rendered textual glyphs. Unlike linguistic or structural text based steganography, the proposed method operates exclusively after font rasterization, modifying only the bitmap produced by a deterministic text rendering pipeline. Each glyph functions as a covert encoding unit, where a payload value is expressed through the cardinality of minimally perturbed interior ink pixels. These minimal intensity increments remain visually imperceptible while forming a stable and decodable signal. The framework is demonstrated for text to text embedding and generalized to multimodal inputs by normalizing image intensities, audio derived scalar features, and video frame values into bounded integer sequences distributed across glyphs. Decoding is achieved by re-rasterizing the cover text, subtracting canonical glyph rasters, and recovering payload values via pixel count analysis. The approach is computationally lightweight, and grounded in deterministic raster behavior, enabling ordinary text to serve as a visually covert medium for multimodal data embedding.

Retrieval-Augmented Code Generation (RACG) is increasingly adopted to enhance Large Language Models for software development, yet its security implications remain dangerously underexplored. This paper conducts the first systematic exploration of a critical and stealthy threat: backdoor attacks targeting the retriever component, which represents a significant supply-chain vulnerability. It is infeasible to assess this threat realistically, as existing attack methods are either too ineffective to pose a real danger or are easily detected by state-of-the-art defense mechanisms spanning both latent-space analysis and token-level inspection, which achieve consistently high detection rates. To overcome this barrier and enable a realistic analysis, we first developed VenomRACG, a new class of potent and stealthy attack that serves as a vehicle for our investigation. Its design makes poisoned samples statistically indistinguishable from benign code, allowing the attack to consistently maintain low detectability across all evaluated defense mechanisms. Armed with this capability, our exploration reveals a severe vulnerability: by injecting vulnerable code equivalent to only 0.05% of the entire knowledge base size, an attacker can successfully manipulate the backdoored retriever to rank the vulnerable code in its top-5 results in 51.29% of cases. This translates to severe downstream harm, causing models like GPT-4o to generate vulnerable code in over 40% of targeted scenarios, while leaving the system's general performance intact. Our findings establish that retriever backdooring is not a theoretical concern but a practical threat to the software development ecosystem that current defenses are blind to, highlighting the urgent need for robust security measures.

Passwordless authentication has revolutionized the way we authenticate across various websites and services. FIDO2 Passkeys, is one of the most-widely adopted standards of passwordless authentication that promises phishing-resistance. However, like any other authentication system, passkeys require the user details to be saved on a centralized server, also known as Relying Party (RP) Server. This has led users to create a new passkey for every new online account. While this just works for a limited number of online accounts, the limited storage space of secure storage modules like TPM or a physical security key limits the number of passkeys a user can have. For example, Yubico Yubikey 5 (firmware 5.0 - 5.6) offers to store only 25 passkeys, while firmware 5.7+ allows to store upto 100 [1]. To overcome this problem, one of the widely adopted approaches is to use Federated Authentication with Single Sign On (SSO). This allows the user to create a passkey for the Identity Provider (IdP) and use the IdP to authenticate to all service providers. This proves to be a significant privacy risk since the IdP can potentially track users across different services. To overcome these limitations, this paper introduces a novel standard 'Verifiable Passkey' that allows the user to use Passkeys created for a Verifiable Credential issuer across any platform without risking privacy or user tracking.

The integration of non-terrestrial networks (NTNs) into 6G systems is crucial for achieving seamless global coverage, particularly in underserved and disaster-prone regions. Among NTN platforms, unmanned aerial vehicles (UAVs) are especially promising due to their rapid deployability. However, this shift from fixed, wired base stations (BSs) to mobile, wireless, energy-constrained UAV-BSs introduces unique security challenges. Their central role in emergency communications makes them attractive candidates for emergency alert spoofing. Their limited computing and energy resources make them more vulnerable to denial-of-service (DoS) attacks, and their dependence on wireless backhaul links and GNSS navigation exposes them to jamming, interception, and spoofing. Furthermore, UAV mobility opens new attack vectors such as malicious handover manipulation. This paper identifies several attack surfaces of UAV-BS systems and outlines principles for mitigating their threats.

With the rapid development of quantum computing, classical cryptography systems are increasingly vulnerable to security threats, thereby highlighting the urgency of constructing architectures that are resilient to quantum computing attacks. While Quantum Key Distribution (QKD) offers security with information-theoretic guarantees, its relatively low key generation rate necessitates integration with classical cryptographic techniques, particularly block ciphers such as AES and SM4, to facilitate practical applications. However, when a single QKD-key is employed to encrypt multiple data blocks, the reduction in cryptographic security strength has not yet been quantitatively analyzed. In this work, we focus on the security strength in the application scenario where QKD is combined with block ciphers. We propose a quantitative evaluation method for the security benefits of the QKD-key renewal period, aiming to provide a precise measure of the cryptographic security strength in such hybrid systems. Our method is based on concrete security paradigm of block cipher modes of operation. We demonstrate that under practical security level requirements, for files consisting of specific blocks, rekeying k times can provide an additional log2(k) to 2log2(k) bits of security. Our research offers a novel perspective on balancing the security and efficiency of QKD-based encryption.