Papers

With the rise of software-defined vehicles (SDVs), where software governs most vehicle functions alongside enhanced connectivity, the need for secure software updates has become increasingly critical. Software vulnerabilities can severely impact safety, the economy, and society. In response to this challenge, Strandberg et al. [escar Europe, 2021] introduced the Unified Software Update Framework (UniSUF), designed to provide a secure update framework that integrates seamlessly with existing vehicular infrastructures. Although UniSUF has previously been evaluated regarding cybersecurity, these assessments have not employed formal verification methods. To bridge this gap, we perform a formal security analysis of UniSUF. We model UniSUF's architecture and assumptions to reflect real-world automotive systems and develop a ProVerif-based framework that formally verifies UniSUF's compliance with essential security requirements - confidentiality, integrity, authenticity, freshness, order, and liveness - demonstrating their satisfiability through symbolic execution. Our results demonstrate that UniSUF adheres to the specified security guarantees, ensuring the correctness and reliability of its security framework.

Quantum Key Distribution (QKD) networks enable unconditionally secure key exchange using quantum mechanical principles. However, routing cryptographic keys across multi-hop quantum networks introduces challenges unique to quantum communication. This survey analyzes and classifies 26 routing strategies proposed between 2013 and 2024 for terrestrial, satellite, and hybrid QKD infrastructures. Dynamic, key-aware routing algorithms have been shown to reduce service rejection rates by 25-40% compared to static shortest-path methods by incorporating real-time key pool availability and link error rates. Multi-path strategies improve resilience against trusted-node compromise by distributing keys across disjoint routes, albeit with an increase in key consumption of 30-60%. SDN-based orchestration frameworks emerge as essential enablers of flexible, QoS-aware routing for hybrid networks integrating terrestrial fibers and intermittent satellite links. Critical research directions include reducing dependence on trusted nodes through quantum repeaters or post-quantum cryptography, developing standardized key management interfaces for interoperability, and adopting AI-driven predictive routing for adapting to fluctuating network states. This comprehensive synthesis of routing algorithms, performance trade-offs, and unresolved issues offers practical guidance for engineering secure, adaptable, and large-scale QKD networks capable of meeting real-world deployment demands.

Existing rug pull detectors assume a simple workflow: the deployer keeps liquidity pool (LP) tokens and performs one or a few large sells (within a day) that collapse the pool and cash out. In practice, however, many real-world exits violate these assumptions by splitting the attack across both time and actor dimensions: attackers break total extraction into many low-impact trades and route proceeds through multiple non-owner addresses, producing low-visibility drains. We formalize this family of attacks as the fragmented rug pull (FRP) and offer a compact recipe for a slow-stewed beef special: (i) keep the lid on (to preserve LP control so on-chain extraction remains feasible), (ii) chop thin slices (to split the total exit volume into many low-impact micro-trades that individually fall below impact thresholds), and (iii) pass the ladle (to delegate sells across multiple wallets so that each participant takes a small share of the extraction). Technically, we define three atomic predicate groups and show that their orthogonal combinations yield evasive strategies overlooked by prior heuristics (USENIX Sec 19, USENIX Sec 23). We validate the model with large-scale measurements. Our corpus contains 303,614 LPs, among which 105,434 are labeled as FRP pools. The labeled subset includes 34,192,767 pool-related transactions and 401,838 inflated-seller wallets, involving 1,501,408 unique interacting addresses. Notably, owner-wallet participation in inflated selling among FRP-flagged LPs has declined substantially (33.1% of cases), indicating a shift in scam behavior: the liquidity drain is no longer held on the owner wallet. We also detected 127,252 wallets acting as serial scammers when repeatedly engaging in inflated selling across multiple FRP LPs. Our empirical findings demonstrate that the evasive strategies we define are widespread and operationally significant.

Phishing websites pose a major cybersecurity threat, exploiting unsuspecting users and causing significant financial and organisational harm. Traditional machine learning approaches for phishing detection often require extensive feature engineering, continuous retraining, and costly infrastructure maintenance. At the same time, proprietary large language models (LLMs) have demonstrated strong performance in phishing-related classification tasks, but their operational costs and reliance on external providers limit their practical adoption in many business environments. This paper investigates the feasibility of small language models (SLMs) for detecting phishing websites using only their raw HTML code. A key advantage of these models is that they can be deployed on local infrastructure, providing organisations with greater control over data and operations. We systematically evaluate 15 commonly used Small Language Models (SLMs), ranging from 1 billion to 70 billion parameters, benchmarking their classification accuracy, computational requirements, and cost-efficiency. Our results highlight the trade-offs between detection performance and resource consumption, demonstrating that while SLMs underperform compared to state-of-the-art proprietary LLMs, they can still provide a viable and scalable alternative to external LLM services. By presenting a comparative analysis of costs and benefits, this work lays the foundation for future research on the adaptation, fine-tuning, and deployment of SLMs in phishing detection systems, aiming to balance security effectiveness and economic practicality.

We study financial transaction confirmation finality in Bitcoin as a function of transaction amount and user risk tolerance. A transaction is recorded in a block on a blockchain. However, a transaction may be revoked due to a fork in the blockchain, the odds of which decrease over time but never reach zero. Therefore, a transaction is considered confirmed if its block is sufficiently deep in the blockchain. This depth is usually set empirically at some fixed number such as six blocks. We analyze forks under varying network delays in simulation and actual Bitcoin data. Based on this analysis, we establish a relationship between block depth and the probability of confirmation revocation due to a fork. We use prospect theory to relate transaction confirmation probability to transaction amount and user risk tolerance.

Blockchain consensus faces a trilemma of security, latency, and decentralization. High-throughput systems often require a reduction in decentralization or robustness against strong adversaries, while highly decentralized and secure systems tend to have lower performance. We present BlueBottle, a two-layer consensus architecture. The core layer, BB-Core, is an n=5f+1 protocol that trades some fault tolerance for a much lower finality latency with a medium-sized core validator set. Our experiments show that BB-Core reduces latency by 20-25% in comparison to Mysticeti. The guard layer, BB-Guard, provides decentralized timestamping, proactive misbehavior detection in BB-Core, and a synchronous recovery path. When it observes equivocations or liveness failures in the core – while tolerating up to f<3n/5 faulty nodes in the primary layer – guard validators disseminate evidence, agree on misbehaving parties for exclusion or slashing, and either restart the core protocol (for liveness violations) or select a canonical fork (for safety violations). Together, these layers enable optimistic sub-second finality at high throughput while maintaining strong safety and liveness under a mild synchrony assumption.

The proliferation of IoT devices in shared, multi-vendor environments like the modern aircraft cabin creates a fundamental conflict between the promise of data collaboration and the risks to passenger privacy, vendor intellectual property (IP), and regulatory compliance. While emerging standards like the Cabin Secure Media-Independent Messaging (CSMIM) protocol provide a secure communication backbone, they do not resolve data governance challenges at the application layer, leaving a privacy gap that impedes trust. This paper proposes and evaluates a framework that closes this gap by integrating a configurable layer of Privacy-Enhancing Technologies (PETs) atop a CSMIM-like architecture. We conduct a rigorous, empirical analysis of two pragmatic PETs: Differential Privacy (DP) for statistical sharing, and an additive secret sharing scheme (ASS) for data obfuscation. Using a high-fidelity testbed with resource-constrained hardware, we quantify the trade-offs between data privacy, utility, and computing performance. Our results demonstrate that the computational overhead of PETs is often negligible compared to inherent network and protocol latencies. We prove that architectural choices, such as on-device versus virtualized processing, have a far greater impact on end-to-end latency and computational performance than the PETs themselves. The findings provide a practical roadmap for system architects to select and configure appropriate PETs, enabling the design of trustworthy collaborative IoT ecosystems in avionics and other critical domains.

The security of future large-scale IoT networks is critically threatened by the “Harvest Now, Decrypt Later” (HNDL) attack paradigm. Securing the massive, long-lived data streams from these systems requires protocols that are both quantum-resistant and highly scalable. Existing solutions are insufficient: post-quantum classical protocols rely on computational assumptions that may not hold for decades, while purely quantum protocols are too resource-intensive for the sheer scale of IoT. This paper introduces the Quantum Anonymous Data Reporting (QADR) protocol, a hybrid framework that provides a theoretical benchmark and high-performance architecture for this challenge, designed for future fully-connected quantum networks. The protocol achieves scalable, quantum-resistant anonymity through a hybrid security model; it leverages information-theoretically secure keys from Quantum Key Distribution (QKD) to seed a quantum-secure pseudorandom function (QS-PRF), grounding its long-term data protection in well-established computational hardness assumptions. We also propose and analyze an automated slot reservation mechanism by making a deliberate trade-off: achieving high performance by accepting a quantifiable information leak during the anonymous slot reservation phase while maintaining strong unlinkability for the final data submission. Our security analysis formally quantifies the anonymity reduction caused by the leak and discusses pathways to fully mitigate it at a significant performance cost. We prove the protocol's critical advantage as a performance benchmark: its primary communication cost scales as $O(n^2)$, a dramatic improvement over quantum-native alternatives ($O(n^4)$), establishing a high-performance goal for future quantum-secured anonymity systems.

The integration of generative artificial intelligence (GenAI) into 6G networks promises substantial performance gains while simultaneously exposing novel security vulnerabilities rooted in multimodal data processing and autonomous reasoning. This article presents a unified perspective on cross-domain vulnerabilities that arise across integrated sensing and communication (ISAC), federated learning (FL), digital twins (DTs), diffusion models (DMs), and large telecommunication models (LTMs). We highlight emerging adversarial agents such as compromised DTs and LTMs that can manipulate both the physical and cognitive layers of 6G systems. To address these risks, we propose an adaptive evolutionary defense (AED) concept that continuously co-evolves with attacks through GenAI-driven simulation and feedback, combining physical-layer protection, secure learning pipelines, and cognitive-layer resilience. A case study using an LLM-based port prediction model for fluid-antenna systems demonstrates the susceptibility of GenAI modules to adversarial perturbations and the effectiveness of the proposed defense concept. Finally, we summarize open challenges and future research directions toward building trustworthy, quantum-resilient, and adaptive GenAI-enabled 6G networks.

Large Language Model (LLM)-based agents with function-calling capabilities are increasingly deployed, but remain vulnerable to Indirect Prompt Injection (IPI) attacks that hijack their tool calls. In response, numerous IPI-centric defense frameworks have emerged. However, these defenses are fragmented, lacking a unified taxonomy and comprehensive evaluation. In this Systematization of Knowledge (SoK), we present the first comprehensive analysis of IPI-centric defense frameworks. We introduce a comprehensive taxonomy of these defenses, classifying them along five dimensions. We then thoroughly assess the security and usability of representative defense frameworks. Through analysis of defensive failures in the assessment, we identify six root causes of defense circumvention. Based on these findings, we design three novel adaptive attacks that significantly improve attack success rates targeting specific frameworks, demonstrating the severity of the flaws in these defenses. Our paper provides a foundation and critical insights for the future development of more secure and usable IPI-centric agent defense frameworks.

The rapid proliferation of Multimodal Large Language Models (MLLMs) has introduced unprecedented security challenges, particularly in phishing detection within academic environments. Academic institutions and researchers are high-value targets, facing dynamic, multilingual, and context-dependent threats that leverage research backgrounds, academic collaborations, and personal information to craft highly tailored attacks. Existing security benchmarks largely rely on datasets that do not incorporate specific academic background information, making them inadequate for capturing the evolving attack patterns and human-centric vulnerability factors specific to academia. To address this gap, we present AdapT-Bench, a unified methodological framework and benchmark suite for systematically evaluating MLLM defense capabilities against dynamic phishing attacks in academic settings.

The AI trustworthiness crisis threatens to derail the artificial intelligence revolution, with regulatory barriers, security vulnerabilities, and accountability gaps preventing deployment in critical domains. Current AI systems operate on opaque data structures that lack the audit trails, provenance tracking, or explainability required by emerging regulations like the EU AI Act. We propose an artifact-centric AI agent paradigm where behavior is driven by persistent, verifiable data artifacts rather than ephemeral tasks, solving the trustworthiness problem at the data architecture level. Central to this approach is the Multimodal Artifact File Format (MAIF), an AI-native container embedding semantic representations, cryptographic provenance, and granular access controls. MAIF transforms data from passive storage into active trust enforcement, making every AI operation inherently auditable. Our production-ready implementation demonstrates ultra-high-speed streaming (2,720.7 MB/s), optimized video processing (1,342 MB/s), and enterprise-grade security. Novel algorithms for cross-modal attention, semantic compression, and cryptographic binding achieve up to 225 compression while maintaining semantic fidelity. Advanced security features include stream-level access control, real-time tamper detection, and behavioral anomaly analysis with minimal overhead. This approach directly addresses the regulatory, security, and accountability challenges preventing AI deployment in sensitive domains, offering a viable path toward trustworthy AI systems at scale.

Efficient zero-knowledge proofs (ZKPs) have been restricted to NP statements so far, whereas they exist for all statements in PSPACE. This work presents the first practical zero-knowledge (ZK) protocols for PSPACE-complete statements by enabling ZK proofs of QBF (Quantified Boolean Formula) evaluation. The core idea is to validate quantified resolution proofs (Q-Res) in ZK. We develop an efficient polynomial encoding of Q-Res proofs, enabling proof validation through low-overhead arithmetic checks. We also design a ZK protocol to prove knowledge of a winning strategy related to the QBF, which is often equally important in practice. We implement our protocols and evaluate them on QBFEVAL. The results show that our protocols can verify 72% of QBF evaluations via Q-Res proof and 82% of instances' winning strategies within 100 seconds, for instances where such proofs or strategies can be obtained.

The shuffle model enhances privacy by anonymizing users' reports through random permutation. This paper presents the first systematic study of the single-message shuffle model from an information-theoretic perspective. We analyze two regimes: the shuffle-only setting, where each user directly submits its message ($Y_i=X_i$), and the shuffle-DP setting, where each user first applies a local $\varepsilon_0$-differentially private mechanism before shuffling ($Y_i=\mathcal{R}(X_i)$). Let $\boldsymbol{Z} = (Y_{\sigma(i)})_i$ denote the shuffled sequence produced by a uniformly random permutation $\sigma$, and let $K = \sigma^{-1}(1)$ represent the position of user 1's message after shuffling. For the shuffle-only setting, we focus on a tractable yet expressive basic configuration, where the target user's message follows $Y_1 \sim P$ and the remaining users' messages are i.i.d.\ samples from $Q$, i.e., $Y_2,\dots,Y_n \sim Q$. We derive asymptotic expressions for the mutual information quantities $I(Y_1;\boldsymbol{Z})$ and $I(K;\boldsymbol{Z})$ as $n \to \infty$, and demonstrate how this analytical framework naturally extends to settings with heterogeneous user distributions. For the shuffle-DP setting, we establish information-theoretic upper bounds on total information leakage. When each user applies an $\varepsilon_0$-DP mechanism, the overall leakage satisfies $I(K; \boldsymbol{Z}) \le 2\varepsilon_0$ and $I(X_1; \boldsymbol{Z}\mid (X_i)_{i=2}^n) \le (e^{\varepsilon_0}-1)/(2n) + O(n^{-3/2})$. These results bridge shuffle differential privacy and mutual-information-based privacy.

Recently, the number of malicious open-source packages in package repositories has been increasing dramatically. While major security scanners focus on identifying known Common Vulnerabilities and Exposures (CVEs) in open-source packages, there are very few studies on detecting malicious packages. Malicious open-source package detection typically requires static, dynamic analysis, or both. Dynamic analysis is more effective as it can expose a package's behaviors at runtime. However, current dynamic analysis tools (e.g., ossf's package-analysis) lack an automatic method to differentiate malicious packages from benign packages. In this paper, we propose an approach to extract the features from dynamic analysis (e.g., executed commands) and leverage machine learning techniques to automatically classify packages as benign or malicious. Our evaluation of nearly 2000 packages on npm shows that the machine learning classifier achieves an AUC of 0.91 with a false positive rate of nearly 0%.

Large-scale cyber-physical systems (CPS), such as railway control systems and smart grids, consist of geographically distributed subsystems that are connected via unreliable, asynchronous inter-region networks. Their scale and distribution make them especially vulnerable to faults and attacks. Unfortunately, existing fault-tolerant methods either consume excessive resources or provide only eventual guarantees, making them unsuitable for real-time resource-constrained CPS. We present GeoShield, a resource-efficient solution for defending geo-distributed CPS against Byzantine faults. GeoShield leverages the property that CPS are designed to tolerate brief disruptions and maintain safety, as long as they recover (i.e., resume normal operations or transition to a safe mode) within a bounded amount of time following a fault. Instead of masking faults, it detects them and recovers the system within bounded time, thus guaranteeing safety with much fewer resources. GeoShield introduces protocols for Byzantine fault-resilient network measurement and inter-region omission fault detection that proactively detect malicious message delays, along with recovery mechanisms that guarantee timely recovery while maximizing operational robustness. It is the first bounded-time recovery solution that operates effectively under unreliable networks without relying on trusted hardware. Evaluations using real-world case studies show that it significantly outperforms existing methods in both effectiveness and resource efficiency.

Quantum Machine Learning (QML) integrates quantum computational principles into learning algorithms, offering improved representational capacity and computational efficiency. However, the security and robustness of QML systems remain underexplored, particularly under adversarial conditions. We present the first comprehensive systematization of adversarial robustness in QML, combining conceptual organization with empirical evaluation across black-box, gray-box, and white-box threat models. We implement five representative attacks: a label-flipping poisoning attack under black-box; an encoder-level indiscriminate poisoning attack and a proxy-model clean-label backdoor attack under gray-box; and a circuit-level backdoor attack (QTrojan) and gradient-based evasion attacks (FGSM and PGD) under white-box. We evaluate these attacks using a Quantum Multilayer Perceptron (QMLP) trained on MNIST and AZ-Class across circuit depths of 2, 5, 10, and 50 layers with angle and amplitude encoding schemes. Our evaluations reveal a fundamental accuracy-robustness trade-off. Amplitude encoding achieves the highest clean accuracy (92.6% on MNIST and 67% on AZ-Class) but collapses under adversarial perturbations and depolarizing noise, whereas shallow angle-encoded models remain more stable. QUID is effective under noiseless conditions but weakened by noise, while the proxy-model backdoor persists unless the circuit itself is overwhelmed. Furthermore, the circuit-level backdoor fails in the multi-class setting, indicating a scalability limitation. Finally, QMLP models are more robust than Classical Multi-Layer Perceptron (CMLP) models under label-flipping attacks but substantially more vulnerable to gradient-based evasion. We conclude by proposing a threat-aware and noise-resilient framework for secure QML deployment.

This paper proposes an automated LLM-based method to extract and structure vulnerabilities from OpenVAS and Tenable WAS scanner reports, converting unstructured data into a standardized format for risk management. In an evaluation using a report with 34 vulnerabilities, GPT-4.1 and DeepSeek achieved the highest similarity to the baseline (ROUGE-L greater than 0.7). The method demonstrates feasibility in transforming complex reports into usable datasets, enabling effective prioritization and future anonymization of sensitive data.

This work presents AnonLFI 2.0, a modular pseudonymization framework for CSIRTs that uses HMAC SHA256 to generate strong and reversible pseudonyms, preserves XML and JSON structures, and integrates OCR and technical recognizers for PII and security artifacts. In two case studies involving OCR applied to PDF documents and an OpenVAS XML report, the system achieved perfect precision and F1 scores of 76.5 and 92.13, demonstrating its effectiveness for securely preparing complex cybersecurity datasets.

Machine learning (ML)-based malware detectors degrade over time as concept drift introduces new and evolving families unseen during training. Retraining is limited by the cost and time of manual labeling or sandbox analysis. Existing approaches mitigate this via drift detection and selective labeling, but fully label-free adaptation remains largely unexplored. Recent self-training methods use a previously trained model to generate pseudo-labels for unlabeled data and then train a new model on these labels. The unlabeled data are used only for inference and do not participate in training the earlier model. We argue that these unlabeled samples still carry valuable information that can be leveraged when incorporated appropriately into training. This paper introduces LFreeDA, an end-to-end framework that adapts malware classifiers to drift without manual labeling or drift detection. LFreeDA first performs unsupervised domain adaptation on malware images, jointly training on labeled and unlabeled samples to infer pseudo-labels and prune noisy ones. It then adapts a classifier on CFG representations using the labeled and selected pseudo-labeled data, leveraging the scalability of images for pseudo-labeling and the richer semantics of CFGs for final adaptation. Evaluations on the real-world MB-24+ dataset show that LFreeDA improves accuracy by up to 12.6% and F1 by 11.1% over no-adaptation lower bounds, and is only 4% and 3.4% below fully supervised upper bounds in accuracy and F1, respectively. It also matches the performance of state-of-the-art methods provided with ground truth labels for 300 target samples. Additional results on two controlled-drift benchmarks further confirm that LFreeDA maintains malware detection performance as malware evolves without human labeling.

Large Language Models (LLMs) increasingly use persistent memory from past interactions to enhance personalization and task performance. However, this memory introduces critical risks when sensitive information is revealed in inappropriate contexts. We present CIMemories, a benchmark for evaluating whether LLMs appropriately control information flow from memory based on task context. CIMemories uses synthetic user profiles with over 100 attributes per user, paired with diverse task contexts in which each attribute may be essential for some tasks but inappropriate for others. Our evaluation reveals that frontier models exhibit up to 69% attribute-level violations (leaking information inappropriately), with lower violation rates often coming at the cost of task utility. Violations accumulate across both tasks and runs: as usage increases from 1 to 40 tasks, GPT-5's violations rise from 0.1% to 9.6%, reaching 25.1% when the same prompt is executed 5 times, revealing arbitrary and unstable behavior in which models leak different attributes for identical prompts. Privacy-conscious prompting does not solve this - models overgeneralize, sharing everything or nothing rather than making nuanced, context-dependent decisions. These findings reveal fundamental limitations that require contextually aware reasoning capabilities, not just better prompting or scaling.

In this study, we evaluate open-source models for security incident classification, comparing them with proprietary models. We utilize a dataset of anonymized real incidents, categorized according to the NIST SP 800-61r3 taxonomy and processed using five prompt-engineering techniques (PHP, SHP, HTP, PRP, and ZSL). The results indicate that, although proprietary models still exhibit higher accuracy, locally deployed open-source models provide advantages in privacy, cost-effectiveness, and data sovereignty.

To autonomously control vehicles, driving agents use outputs from a combination of machine-learning (ML) models, controller logic, and custom modules. Although numerous prior works have shown that adversarial examples can mislead ML models used in autonomous driving contexts, it remains unclear if these attacks are effective at producing harmful driving actions for various agents, environments, and scenarios. To assess the risk of adversarial examples to autonomous driving, we evaluate attacks against a variety of driving agents, rather than against ML models in isolation. To support this evaluation, we leverage CARLA, an urban driving simulator, to create and evaluate adversarial examples. We create adversarial patches designed to stop or steer driving agents, stream them into the CARLA simulator at runtime, and evaluate them against agents from the CARLA Leaderboard, a public repository of best-performing autonomous driving agents from an annual research competition. Unlike prior work, we evaluate attacks against autonomous driving systems without creating or modifying any driving-agent code and against all parts of the agent included with the ML model. We perform a case-study investigation of two attack strategies against three open-source driving agents from the CARLA Leaderboard across multiple driving scenarios, lighting conditions, and locations. Interestingly, we show that, although some attacks can successfully mislead ML models into predicting erroneous stopping or steering commands, some driving agents use modules, such as PID control or GPS-based rules, that can overrule attacker-manipulated predictions from ML models.

Attack trees (ATs) are popular graphical models for reasoning about the security of complex systems, allowing for the quantification of risk through so-called AT metrics. A large variety of different such AT metrics have been proposed, and despite their wide-spread practical use, no systematic treatment of attack tree metrics so far is fully satisfactory. Existing approaches either fail to include important metrics, or they are too general to provide a useful systematic way for defining concrete AT metrics, giving only an abstract characterisation of their behaviour. We solve this problem by developing a compositional theory of ATs and their functorial semantics based on gs-monoidal categories. Viewing attack trees as string diagrams, we show that components of ATs form a channel category, a particular type of gs-monoidal category. AT metrics then correspond to functors of channel categories. This characterisation is both general enough to include all common AT metrics, and concrete enough to define AT metrics by their logical structure.

Federated learning (FL) enables collaborative model training while preserving data privacy. However, it remains vulnerable to malicious clients who compromise model integrity through Byzantine attacks, data poisoning, or adaptive adversarial behaviors. Existing defense mechanisms rely on static thresholds and binary classification, failing to adapt to evolving client behaviors in real-world deployments. We propose FLARE, an adaptive reputation-based framework that transforms client reliability assessment from binary decisions to a continuous, multi-dimensional trust evaluation. FLARE integrates: (i) a multi-dimensional reputation score capturing performance consistency, statistical anomaly indicators, and temporal behavior, (ii) a self-calibrating adaptive threshold mechanism that adjusts security strictness based on model convergence and recent attack intensity, (iii) reputation-weighted aggregation with soft exclusion to proportionally limit suspicious contributions rather than eliminating clients outright, and (iv) a Local Differential Privacy (LDP) mechanism enabling reputation scoring on privatized client updates. We further introduce a highly evasive Statistical Mimicry (SM) attack, a benchmark adversary that blends honest gradients with synthetic perturbations and persistent drift to remain undetected by traditional filters. Extensive experiments with 100 clients on MNIST, CIFAR-10, and SVHN demonstrate that FLARE maintains high model accuracy and converges faster than state-of-the-art Byzantine-robust methods under diverse attack types, including label flipping, gradient scaling, adaptive attacks, ALIE, and SM. FLARE improves robustness by up to 16% and preserves model convergence within 30% of the non-attacked baseline, while achieving strong malicious-client detection performance with minimal computational overhead. https://github.com/Anonymous0-0paper/FLARE

Mobile Web3 faces catastrophic retention (< 5%) yielding effective acquisition costs of \$500 - \$1,000 per retained user. Existing solutions force an impossible tradeoff: embedded wallets achieve moderate usability but suffer inherent click-jacking vulnerabilities; app wallets maintain security at the cost of 2 - 3% retention due to download friction and context-switching penalties. We present SecureSign, a PWA-based architecture that adapts desktop browser extension security to mobile via EIP-6963 provider sandboxing. SecureSign isolates dApp execution in iframes within a trusted parent application, achieving click-jacking immunity and transaction integrity while enabling native mobile capabilities (push notifications, home screen installation, zero context-switching). Our drop-in SDK requires no codebase changes for existing Web3 applications. Threat model analysis demonstrates immunity to click-jacking, overlay, and skimming attacks while maintaining wallet interoperability across dApps.

Modern deepfakes evade detection by leaving subtle, domain-speci c artifacts that single branch networks miss. ForensicFlow addresses this by fusing evidence across three forensic dimensions: global visual inconsistencies (via ConvNeXt-tiny), ne-grained texture anomalies (via Swin Transformer-tiny), and spectral noise patterns (via CNN with channel attention). Our attention-based temporal pooling dynamically prioritizes high-evidence frames, while adaptive fusion weights each branch according to forgery type. Trained on CelebDF(v2) with Focal Loss, the model achieves AUC 0.9752, F1 0.9408, and accuracy 0.9208 out performing single-stream detectors. Ablation studies con rm branch synergy, and Grad-CAM visualizations validate focus on genuine manipulation regions (e.g., facial boundaries). This multi-domain fusion strategy establishes robustness against increasingly sophisticated forgeries.

We show that an i.i.d.\ binary source sequence $X_1,\ldots,X_n$ can be losslessly compressed at any rate above entropy while ensuring that the decoding of any $X_i$ reveals no information about the remaining symbols $\{X_j : j \neq i\}$. This problem reduces to a marginal consistency problem induced by the simultaneous privacy and reliability constraints. To address it, we develop a technique based on a geometric representation of codeword distributions, which may be of independent interest.

U-shaped split federated learning (U-SFL) enables resource-constrained Internet of Things devices to collaboratively train models with an edge server while retaining raw data and labels locally. We consider model-service deployments in which a provider supplies proprietary models whose client-side segments reside on participating devices, creating risks of unauthorized copying and redistribution. Protecting these segments is challenging because the server cannot directly access client data, labels, or model parameters, and potentially malicious clients may refuse to perform watermark embedding. To address this challenge, we propose Sigil, a server-enforced watermarking framework for U-SFL. Sigil defines a secret watermark constraint in the server-visible activation space and embeds the watermark into client-side models by injecting a watermark gradient into the gradients returned during training. This mechanism requires neither access to clients' raw data and labels nor client-side watermarking operations. To limit interference with the main task and reduce detectability by gradient anomaly detectors, Sigil adaptively clips the watermark gradient relative to the main-task gradient. Experiments on two datasets and four model architectures demonstrate high watermark detection rates with limited impact on task accuracy, robustness against the evaluated removal attacks, and stealthiness against the evaluated gradient anomaly detector.

Large models adaptation through Federated Learning (FL) addresses a wide range of use cases and is enabled by Parameter-Efficient Fine-Tuning techniques such as Low-Rank Adaptation (LoRA). However, this distributed learning paradigm faces several security threats, particularly to its integrity, such as backdoor attacks that aim to inject malicious behavior during the local training steps of certain clients. We present the first analysis of the influence of LoRA on state-of-the-art backdoor attacks targeting model adaptation in FL. Specifically, we focus on backdoor lifespan, a critical characteristic in FL, that can vary depending on the attack scenario and the attacker's ability to effectively inject the backdoor. A key finding in our experiments is that for an optimally injected backdoor, the backdoor persistence after the attack is longer when the LoRA's rank is lower. Importantly, our work highlights evaluation issues of backdoor attacks against FL and contributes to the development of more robust and fair evaluations of backdoor attacks, enhancing the reliability of risk assessments for critical FL systems. Our code is publicly available.