Papers

Continual learning (CL) models often use experience replay to reduce catastrophic forgetting, but their robustness to replay sampling interference remains underexplored. Existing CL attacks alter inputs or training pipelines (poisoning/backdoors) and rarely include explicit auditable constraints, limiting realism. Here, auditability means a monitor can verify compliance from sampler-visible telemetry - e.g., logged replay index/label statistics - by checking that the realized replay class histogram stays close to a nominal baseline and that replay rate is unchanged per batch and/or over a rolling window. We study a limited-privilege insider who controls only replay index selection, not pixels, labels, or model parameters, while staying within auditable limits such as queue priorities. We introduce Amnesia, a replay composition attack that maximizes degradation under two budgets: a visibility budget delta bounding the TV/KL divergence from a nominal class histogram p0, and a mass budget f fixing the replay rate. Amnesia has two steps: (i) compute lightweight class utilities, such as EMA loss or confidence, to tilt p0 toward harmful classes; and (ii) project the tilt back into the delta-ball using efficient KL (exponential tilt) or TV (balanced mass redistribution) optimizers. A windowed scheduler enforces rolling audits. Across challenging CL benchmarks and strong replay baselines, Amnesia consistently lowers final accuracy (ACC) and worsens backward transfer (-BWT). The KL variant delivers high impact while remaining largely undetected under multiple audit schemes, including per-batch and rolling-window checks. The TV variant is more damaging but easier to detect, especially under tight per-class constraints. These results expose index-only replay control as a practical, auditable threat surface in CL systems and establish a principled impact-visibility trade-off.

Vision-Language Agentic Systems (VLAS) connect visual perception to planning, tool use, and physical actions. This means backdoor-type triggers can propagate through both decision pipelines and their connected interfaces, thus making visual backdoors a system-level threat. Current evaluations on such backdoors focus on clean accuracy and attack success rate (ASR), metrics that capture whether a trigger works, but not whether an attack is actually "precise" – i.e. whether it triggers hidden behaviors only when intended. In this work, we formalize the failure of trigger precision as "trigger leakage": inputs that are visually or semantically close to the intended trigger and therefore inadvertently activate the attacker-specified behavior. To quantify this leakage, we introduce Neighbor Leakage Rate (NLR). Our experiments show that at a 3% poisoning ratio, icon and text triggers remain robust to common visual transformations, but their neighboring variants leak heavily, with NLR reaching 0.996 (icon) and 0.944 (text). Using textual triggers as a controlled probe, we show that standard fine-tuning learns a broad activation region rather than an exact trigger condition, causing neighboring strings to invoke the malicious behavior even when the exact trigger is absent. Adding edit-distance-one hard-negative samples during training substantially narrows this activation region and reduces leakage, including in image-editing and embodied-manipulation workflows, where leaked triggers can propagate into executable programs and action sequences.

Model merging (MM) has gained significant attention as a cost-effective approach to integrate multiple task-specific models into a unified model. However, recent work reveals that MM is highly susceptible to backdoor attacks. Existing defenses based on task arithmetic often fail to eliminate backdoors without substantially degrading clean-task performance, owing to their reliance on direct parameter-space editing. To address this gap, we propose Linear Feature Path Minimization (LFPM), a backdoor mitigation framework for model merging, which introduces an anti-backdoor task vector into the backdoored merged model. Unlike prior approaches, LFPM formulates the backdoor robustness of the merged model from a unified feature-space perspective under the Cross-Task Linearity (CTL) framework, which leverages the approximate linearity of features across tasks. This perspective guides the optimization of the anti-backdoor task to suppress backdoors while preserving clean-task performance. Furthermore, we introduce an effective optimization mechanism based on gradient accumulation and loss path-integral, ensuring robust backdoor suppression along the interpolation path. Extensive experiments demonstrate that LFPM consistently exhibits strong robustness against backdoor attacks in both full fine-tuning and Parameter-Efficient Fine-Tuning (PEFT) settings.

LLM-based multi-agent systems (MAS) solve complex tasks through inter-agent collaboration, but their communication-driven nature also allows security risks to spread across agents and trigger system-wide failures. Existing MAS defenses mainly follow a reactive paradigm after execution by detecting and isolating harmful agents, which may cause irreversible damage and degrade collaborative utility. To address this, we propose a proactive defense framework for MAS security, namely a Simulation-aware Interception Guard (SAIGuard). SAIGuard performs communication-state simulation over the MAS interaction graph, estimates the impact of incoming messages on local agent states and the global MAS state, and detects risky messages via reconstruction deviations from benign communication patterns. Instead of isolating agents, SAIGuard sanitizes or regenerates suspicious messages before it propagation into system. Experiments across diverse topologies and attack scenarios show that SAIGuard reduces attack success rates while maintaining MAS utility, outperforming reactive defenses.

As maritime operations increasingly depend on interconnected digital ecosystems, cyber incidents can propagate across maritime networks and degrade critical services. Strengthening strategic Cyber Situational Awareness (CSA) therefore requires training mechanisms that expose decision-makers to evolving attack dynamics, constrained resources, and the need to align actions with incident-response procedures. This paper introduces MARCIM-WG, a learning-oriented maritime cyberdefense wargame designed following the NATO wargaming methodology and implemented as a hybrid tabletop experience combining a physical board (tokens, indicators, and special cards) with analytically-assisted adjudication supported by a computational simulation model. The proposal is specified through High-Level Design (HLD) and Low-Level Design (LLD) specifications and instantiated in a fictional maritime cyber crisis scenario to enable structured decision cycles, friction, and measurable consequences. Validation combines (i) an operational scenario-based assessment under three configurations (pessimistic, neutral/most likely, optimistic) to verify decision sensitivity and outcome coherence, and (ii) a CSA competency and learning-outcome evaluation using a comparative design against an equivalent control group. Results show a +34.0 percentage-point improvement in the intervention group, with the largest gains in comprehension-related competencies.

Operational organizations increasingly require Cyber Situational Awareness (CySA) capabilities that go beyond isolated technical alerts, providing mission-relevant artefacts that can be embedded into heterogeneous toolchains and cyber security or cyber defense processes. ECYSAP EYE addresses this need through an adoption-oriented System-of-Systems (SoS) architecture centered on seven groups of mission-focused artefacts: the Recognized Cyberspace Picture (RCyP), Cyber Situational Reports (CySRs), the What-If Analysis Report (WIAR), Option Recommendations (OPRE), an operator Dashboard/HMI (DSH), Action Enforcement (AE), and After-Action Reports (AAR). The ECYSAP EYE architecture structures the transition from perception (full-spectrum RCyP views), to decision-oriented reasoning (WIAR/CySRs/OPRE), and to operational execution and learning (DSH/AE/AAR), with explicit integration surfaces that support incremental deployment and validation. This paper presents this innovative project from a technology transfer perspective, summarizing the updated architecture, the functional role of seven groups of artefacts, and the expected impact of cyber situations on the decision-making process in the context of a mission planning and execution.

Large language model (LLM) agents increasingly act on a user's behalf – reading personal files, calling tools, transacting with external services – possibly leaking personally identifiable information (PII) across trust boundaries at every step. Privacy here is a property not of a single output but of an entire trajectory, and three properties make it hard: leakage is cumulative, as individually innocuous releases accumulate across honest-but-curious or colluding sinks into inferences about a protected secret; bidirectional, as a malicious observation can inject instructions that turn the agent's own reasoning model against the user; and task-dependent, as the same field is necessary for one recipient yet gratuitous for another. Per-release contextual-integrity filters, information-flow controls, and posterior-leakage monitors each address part of this but none controls cumulative, inference-based leakage at runtime. We recast agent privacy as posterior-risk control and present OCELOT, a runtime mediator that budgets how much an adversary's belief about a secret may improve across a trajectory, rather than filtering outputs. Its mechanism, Witness-Verified Declassification, separates judgment from trust: an untrusted, locally fine-tuned defender model inspects each candidate release and emits structured evidence – labeled atoms and proposed declassification operators – which a deterministic verifier audits, charging a certified min-entropy cost for the chosen variant and authorizing the least-disclosing useful release under a sink-trust-weighted budget recorded on a tamper-evident ledger. Across diverse agent benchmarks and recent defenses, OCELOT attains significantly lower leakage at higher task utility, resists adaptive injection, jailbreak, cumulative inference, and sink collusion, and adds only modest overhead.

Enterprise security was built to govern data boundaries: the protected surface was data at rest and in transit, and the controls – access control, data-loss prevention, perimeter inspection – governed crossings of that boundary. Production AI agents dissolve this assumption. An agent reads context, calls tools, invokes connectors, and modifies systems of record on an enterprise's behalf, so risk moves inside the workflow, into sequences of individually-permitted actions that may transform a business process no one authorized. Existing policy engines do not extend to this regime: they evaluate request-time decisions against atomic principals, where agentic systems require stateful evaluation against composite principals whose authority attenuates through delegation chains. We present a reference architecture for the runtime governance of production agents, built from four composable primitives: a five-plane decomposition (a reasoning plane that adjudicates intent, and four enforcement planes – network, identity, endpoint, data – that realize the decision), stop-anywhere mediation, composite principals with capability attenuation, and audit as a structured evidence substrate. We define a taxonomy of six interruption primitives that generalize allow and deny, state and argue for four correctness invariants, and demonstrate the foreclosure of seven production-agent threats across five concrete workflows. A reference implementation of the policy-engine core supplies measured evidence: attenuation correctness and evidence reconstructability hold on every trial, adjudication runs in single-digit microseconds, and the audit substrate's tamper-evidence behaves exactly as designed. We are explicit about scope: the architecture governs delegated action, not model behavior, and a full-system evaluation against a live agent benchmark is the invited next step.

Agent memory is moving to graphs, and the provenance defenses now being built for it all check one thing: the provenance of the records an agent retrieves. We show that this entire class of defense is blind by construction. A long-term graph memory runs a global selection step over writable graph structure, so structure that an untrusted principal writes changes which authenticated facts are selected while the cited evidence stays fully authenticated; faithful information-flow control (IFC), checking the provenance of what the reader uses (all of it authenticated), makes the byte-identical decision to no defense at all, across document-QA substrates and real multi-session agent memory. In the most consequential instance, a no-source structural write silently misdirects $28$ irreversible ledger transfers over $499$ live actions: faithful IFC permits every one, and \authselect\ prevents every one. We then characterize exactly which memories are exposed: a selector admits the channel when its structural term can reallocate an $\Omega(1)$ share of top-$k$ membership past a selected fact's margin. Personalized PageRank can, since a sourceless write reroutes conserved random-walk mass; a content-fixed reranker cannot, and Graphiti's node-distance, which leans on structure more than PageRank does, stays immune. Reallocatability, not reliance, is the predictor. We prove the immune case in general and the open case under a chokepoint condition we verify. Closing the channel forces any provenance defense to recompute selection on the authenticated subgraph, which is what \authselect\ does, at zero over-block and $2$–$3\%$ latency.

Cache partitioning is among the strongest structural defenses against eviction-based cache side channels, yet a decade-old design issue has blocked its widespread deployment in secure shared-OS settings. The issue is that write-shared coherence collapses under strict partitioning. We present SCP (Secure and Coherent Partitioning), which combines strict eviction isolation with write-shared coherence by partitioning only the tags, sharing a single data pool, and sizing the data pool so capacity-driven cross-partition eviction cannot occur. Timing obfuscation extends protections to the inter-partition lookup path. Coherence-based leakage on shared-writeable lines is mitigated by routing those writes through to the LLC once a leakage threshold is crossed, which makes attacker write probe latency independent of victim activity. Using gem5 for implementation, SCP mitigates Prime+Probe and Flush+Reload, which are the basis for more sophisticated cache attacks. We also demonstrate that a shared-writeable-line attack is mitigated. All these attacks yield results no better than random guessing. SCP's hardware cost is a modest +2.8% LLC SRAM. Performance matches DAWG within 0.3% IPC on the SPEC CPU2017 benchmarks that we evaluated. Sharing-intensive microbenchmarks demonstrate a tunable security-performance tradeoff based on a system-specified leakage threshold.

Gradient-based adversarial attacks remain a dominant threat to deep neural networks (DNNs), as they exploit gradient information to efficiently optimize adversarial perturbations. To address this, we investigate whether reinforcement learning (RL) training can disrupt the gradient structure used by attackers by training image classifiers with policy-gradient objectives and epsilon-greedy exploration. Through systematic experiments across CIFAR-10, CIFAR-100, and ImageNet-100 with multiple architectures, we find that RL-trained classifiers significantly disrupt gradient-based adversarial optimization. To explain this, we conduct a comprehensive mechanism analysis using loss landscape visualization, static and dynamic gradient indicators, and predictive entropy. Our analysis reveals that RL acts as an implicit regularizer, producing models with highly unstable gradient directions and smaller gradient magnitudes. This combination makes each PGD step both unreliable in direction and limited in magnitude, causing gradient-based attacks to fail within practical iteration budgets. We further show that combining RL with adversarial training (RL-adv) provides a dual-layer defense operating at two complementary levels: RL degrades gradient information available to attackers (gradient-level defense), while adversarial training strengthens decision boundaries (boundary-level defense). RL-adv achieves the highest robustness across all major attack types evaluated, including gradient-based (PGD, AutoAttack), transfer-based, and query-based attacks, outperforming SL-adv by a significant margin. These findings identify RL-induced gradient disruption as a complementary robustness mechanism and motivate future research on hybrid SL-RL training schedules that combine SL's efficiency with RL's gradient-regularization properties.

Smart cities rely on interconnected cyber-physical systems that integrate sensors, IoT devices, cloud platforms, and AI-driven services and decision-making. While these systems enhance city services, they also introduce complex cybersecurity challenges due to their large attack surfaces, heterogeneous data flows, and evolving threat vectors. Developing and validating cybersecurity tools for smart cities requires high-quality datasets that accurately represent real operational conditions. However, real-world datasets are often incomplete, contain privacy-sensitive data, are difficult to access, or lack sufficient malicious activity to support tool development. This research addresses this critical gap by proposing an AI-based synthetic data generation (SDG) framework designed specifically for smart city cybersecurity research. The proposed framework leverages generative artificial intelligence models to produce high-fidelity synthetic cybersecurity datasets that replicate realistic device behaviors, network interactions, and cyber-attack scenarios. The synthetic datasets are evaluated for conformity to protocol standards, statistical similarity to original datasets, and utility in common security tools. The resulting synthetic data generation framework and evaluation metrics are expected to advance smart city cybersecurity by enabling researchers to model threats more effectively and evaluate defensive techniques more comprehensively to better protect critical smart city infrastructures.

The rapid integration of large language models (LLMs) into mobile applications has introduced a new class of credential security risk: leaked credentials that grant unauthorized access to LLM inference services, causing financial damage to developers. Prior work on credential leakage has focused primarily on Android apps; to date, no empirical study has systematically investigated LLM API key leakage in iOS applications. We present the first in-depth empirical study of API key leakage in LLM-integrated apps. We construct a high-quality dataset of 444 iOS applications, filtered from 1092 candidates through a standardized process, and develop LLMKeyLens, a dynamic analysis framework that detects LLM API key leakage via traffic interception, provider-specific key extraction, and active validity confirmation, requiring neither source code access nor binary decryption. Our analysis reveals that 282 applications expose exploitable LLM API credentials in network traffic, spanning at least ten providers. We identify three leakage patterns: JWT-based token leakage (48%), unauthenticated backend proxy access (33%), and plaintext API key transmission (19%). To assess remediation, we re-analyzed the same 282 vulnerable applications three months after responsible disclosure; only 28% had remediated the reported vulnerability, while 72% remained exploitable, with persistent issues stemming from unauthenticated backends and broken JWT implementations. Our findings show that LLM API key leakage is both prevalent and persistent in the iOS ecosystem, exposing a systemic gap between developer practice and secure integration principles, and suggest that secure LLM integration requires not only developer awareness but also explicit security guidance from providers and platform-level enforcement.

Network Intrusion Detection Systems (NIDS) heavily utlize Machine Learning (ML) but ML models can be manipulated via adversarial attacks. These attacks add carefully crafted perturbations to network traffic data that leads to misclassifications. While prior work has demonstrated adversarial vulnerabilities in isolated settings, systematic cross-architecture as well as class and category of attack based comparisons under controlled attack conditions remain limited, leaving practitioners without clear guidance on which models to deploy in adversarial environments. This paper asks a simple question: what type of classifier architectures actually hold up when attackers try to manipulate the systems? We put three popular architectures through their paces: a 1D Convolutional Neural Network, a Long Short-Term Memory (LSTM) network, and a Random Forest (RF) ensemble. Using the ACI-IoT-2023 dataset (over 1.2 million samples spanning 12 attack types), we subject each model with FGSM and PGD adversarial attacks, which apply gradient-based perturbations in normalized feature space consistent with established adversarial ML evaluation protocols, at perturbation budgets ranging from $\epsilon=0.01$ to $\epsilon=0.1$. Surprisingly, Random Forest achieved near-perfect baseline accuracy (99.98%), yet collapsed catastrophically under attack, dropping 73 percentage points at the smallest perturbation we tested. CNN, on the other hand, retained 95.5% accuracy at $\epsilon=0.01$ and degraded gracefully as perturbations increased. LSTM fell somewhere in between. These findings flip the conventional wisdom where high baseline accuracy means nothing if a model shatters at the first sign of adversarial pressure. For practitioners deploying intrusion detection in adversarial environments, we recommend CNN-based architectures and provide scenario-specific deployment guidance.

Undefined behavior is idiomatic to C and C++ programming; such behavior is a use of an erroneous program construct for which the languages impose no requirements, such as integer overflows. The paper presents an empirical experiment seeking to probe the extent of undefined behavior executing underneath typical desktop use of a Linux distribution. The analysis is based on an undefined behavior sanitizer implemented in a compiler. According to the results, undefined behavior is common. By completing 59 simple experimental tasks, nearly 11 thousand unique undefined behavior warnings were generated by 32 unique programs and libraries written in C or C++. Of these warnings, most were associated with the Mesa graphics library and generated by interacting with graphical user interfaces. Merely logging into the GNOME desktop environment generated over 500 unique warnings. Of all warnings, the clear majority was about virtual table pointers. The associated stack traces were also lengthy in general. With these and other results, the paper contributes to the empirical literature on C and C++.

Fault Injection Attacks (FIAs) are a significant threat to hardware security, capable of compromising systems by inducing malicious faults in computation or storage. Evaluating resilience against such attacks is challenging due to the high cost, complexity, and limited availability of physical fault experiments, particularly during pre-silicon development. Architectural-level simulation offers a developer-oriented, white-box perspective for systematic vulnerability assessment. This paper introduces InjectV, a fault injection attack framework for RISC-V platforms built on the gem5 simulator. InjectV enables precise, guided fault injection at security-critical execution points, such as control-flow decisions, counters, and comparisons, allowing systematic exploration of attack vectors. It currently supports transient fault attacks in registers and memory, broadening its ability to simulate diverse attack scenarios. Experimental results on security benchmarks from the FISSC suite, including hardened variants of the VerifyPIN application, demonstrate InjectV's ability to effectively identify fault-injection points, achieving a 95.8% time-saving advantage over traditional fault injection approaches.

We describe a computational search for quadratic APN (Almost Perfect Nonlinear) functions over $\mathbb{F}_{2^8}$ within a structured algebraic subspace defined by a self-equivalence constraint. The search space is the 40-dimensional $\mathbb{F}_2$-linear subspace $V_A = \{F : F \circ A = A \circ F\}$ for a specific linear automorphism $A$ of order 5 (class index 22 in the taxonomy of Beierle, Brinkmann, and Leander); this subspace was previously reported to contain no APN functions under their recursive tree search method. We combine two phases: (1) random sampling inside $V_A$ via an explicit RREF parameterization to find APN center functions, and (2) Groebner basis computation in Magma over the Boolean polynomial ring to enumerate all APN functions in a 24-dimensional hyperplane through each center. From 428 hyperplane computations (covering 0.65% of the 65,536 total hyperplanes in $V_A$) we obtained 566 quadratic APN functions falling into six CCZ-equivalence classes under the ortho-derivative invariant. Four of these classes, comprising 500 functions, match no entry in the Beierle et al. 2025 database of 3,775,599 quadratic APN functions and no entry in the pre-2020 compilation of 12,921 instances. Two classes (66 functions) are identified as CCZ-equivalents of the Gold functions x^3 and x^9, confirming pipeline correctness. For quadratic APN functions, a signature mismatch rigorously certifies CCZ-inequivalence, by Yoshiara's theorem (CCZ = EA for quadratic APN) together with the ortho-derivative invariant; the absence of a signature match in the above databases therefore constitutes a rigorous proof of CCZ-inequivalence for the new classes. The complete dataset, source code, and verification scripts are publicly available.

Reasoning models deployed as safety monitors exhibit a systematic vulnerability: reasoning-token budget starvation. Adversarial inputs require $3.3\times$ more reasoning tokens than benign inputs to produce valid safety scores ($T_{50,\text{adv}}{=}154$ vs. $T_{50,\text{benign}}{=}46$ for o3), so low-budget deployments silently starve the monitor on exactly the inputs it must catch. This compounds the central failure mode: gradient-based evasion remains the residual threat - template jailbreaks fail at 99%, but GCG-optimized suffixes flip encoder decisions reliably. We systematize a canary construction - score-disagreement monitoring between a targeted and un-targeted classifier - and quantify its reliability under targeted evasion. We derive the exact security boundary - a confidence-gated equilibrium at which a monitor-aware attacker stalls (validated gap $= 1/(2\lambda)$, within 95% CI of theory) - and identify a failure mode in post-shift conformal adaptation. Three contributions. (1) Factorial drift benchmark. A pre-registered 800-cell evaluation ($4$ classifiers $\times$ $5$ shift types $\times$ $20$ seeds $\times$ $2$ windows) reveals detection difficulty is dominated by a classifier$\times$shift interaction ($\eta^2 = 0.185$): encoders detect paraphrase drift in 28 steps but miss adversarial suffixes for 37; decoders show the opposite. (2) Conformal collapse in generative embeddings. Weighted conformal prediction fails on decoder classifiers: logistic density-ratio estimation achieves perfect separability in 3584–4096-dimensional space, clipping all importance weights to zero. Projecting to $\leq$32 dimensions restores coverage (+33pp). (3) Adversarial canary threat model. Across 35 frontier models, a 4-tier threat model yields deployment guarantees ($\geq$71% detection, $<$1.5% FPR at $N{=}1000$).

This paper addresses the challenge of assessing image quality in ID cards in remote verification systems by applying capture-related quality measures from the Open Face Image Quality (OFIQ) standard to ID card images. Our preprocessing pipeline includes corner detection, perspective normalization, and comprehensive foreground masking to ensure accurate and unbiased quality measure computation. We evaluate the effectiveness of these measures by analyzing their correlation with the performance of three presentation attack detection (PAD) algorithms across four diverse ID card datasets, where two datasets contain bona fide, i.e. pristine, images and two contain printed mock ID cards. Our results suggest that quality assessment based on some OFIQ measures can significantly improve PAD performance.

CUDA collective operations often sit on security decision paths: votes accept batches, reductions aggregate evidence, shuffles select representatives, and barriers order checked state before use. Such decisions depend not only on computed values, but also on which lanes are represented, what evidence they contribute, which lane speaks for the group, and which checked state reaches commit. We identify this participation metadata as decision-making non-control data. We define Collective Semantic Corruption (CSC), a non-control-data attack family in which range-valid masks, predicates, source lanes, descriptors, group labels, or epochs cause a CUDA-conforming collective to authorize a decision over the wrong membership, contribution, role, or validation-to-use state. The kernel reaches the intended collective site and executes the expected primitive; the primitive represents the wrong authority set. We model CSC with a site-local participation-authority contract. A protected collective derives, recomputes, checks, or freezes membership, contribution, role, and temporal state before authorization. We evaluate CSC across NVIDIA CUDA collective primitives, trigger channels, compact workload-style kernels, reduced idiom bridges, and admission-guard harnesses. In a CUDA-defined contract-conformance suite spanning the four authority dimensions, corrupted participation metadata causes a trusted-reference mismatch in 102/102 instances, while hardened variants preserve that reference in 102/102. We report 13 synchronization-sensitive instances separately. We then introduce Collective Integrity Contracts (CIC), a wrapper discipline that binds participation metadata before collective use. For CUDA collective decisions, security depends on both the values computed and the participants represented.

Recent CUDA exploitation work shows that GPU memory bugs can escalate into device-side control-flow corruption, as kernels later consume corrupted return continuations, function pointers, dispatch-table entries, or branch targets. For deployed CUDA binaries, the relevant security boundary is executed NVIDIA SASS, after PTX lowering, inlining, ABI decisions, register allocation, spills, predication, and SIMT execution; source- or PTX-level policies do not capture this boundary. We present WarpGuard, to our knowledge the first protected-site CFI system for CUDA device binaries operating on executed SASS. WarpGuard enforces at protected sites: recovered SASS instructions or sequences that consume control-flow state, provide sufficient binary evidence to derive policy, are checked before release, and fail closed on violation. It authenticates backward-edge continuation state for instrumented returns, validates recoverable forward targets per site, and reports fixed-edge, unsupported, profile-excluded, fallback, and no-surface outcomes outside the protected denominator. On 77 CUDA artifacts, WarpGuard classifies 51,621 SASS control-flow sites, including 1,343 returns and 154 supported forward target-set entries, and records 52.2 million dynamic checks. In representative backward- and forward-edge corruption attacks, native execution reaches attacker-selected behavior, detect-only mode records the expected violation, and enforcement fails closed before releasing the invalid protected transfer. Public-code evidence shows that the same SASS consumption patterns occur in real CUDA systems, including runtime dispatch tables, cuFFT callbacks, generated callable tables, and uploaded device-function pointers. WarpGuard delivers auditable protected-site CFI for CUDA SASS and separates dynamic-instrumentation enforcement from callback-free SASS timing and patch-cache feasibility.

European Rail Traffic Management System (ERTMS) is a widely adopted standard unifying train management in the EU. While the standard allows for use cases like fully autonomous driving, cybersecurity has been an afterthought. Risk analysis enables the systematic assessment and prioritization of threats and mitigations. To date, it remains unclear which threats are most significant in ERTMS. This study systematically models components of ERTMS and analyzes their security in light of threats identified in the underlying technologies. The results suggest a concerning state of ERTMS, despite its critical role in railway safety. The use of legacy standards like EuroBalises and GSM-Railway (GSM-R) introduces vulnerabilities that persist across minimal ERTMS implementations, deployments incorporating various optional safety measures, and prospective future evolutions of the system, e.g., adopting Future Railway Mobile Communication System (FRMCS). Fully transitioning to European Train Control System (ETCS) level 2 was identified as the most significant measure for advancing ERTMS cybersecurity. The results indicate that a shift of ERTMS toward security is required to ensure availability and safe operation. While the chosen methodology proved its feasibility and shows remaining weaknesses of ERTMS, future work is needed to develop railway-centric adaptations to improve the quantification and evaluation of the computed risks.

Audio watermarking aims to embed identifiable information into audio while remaining imperceptible. Existing methods adopt high-fidelity, low-energy designs to preserve perceptual quality, but the resulting watermarks lack robustness under suppression by speech reconstruction models. Improving robustness is challenging due to the inherent robustness-fidelity trade-off in existing designs, where increasing watermark energy improves robustness but reduces fidelity. To address this problem, we propose a feature-aligned watermarking method that aligns the watermark with the original speech feature distribution, allowing higher watermark energy to improve robustness while preserving imperceptibility. We use a pretrained speech codec to generate a pseudo-speech watermark and fuse it into the spectrogram of the input audio, with VAD loss and perceptual losses guiding embedding within voiced regions. Experiments show that our method maintains imperceptibility comparable to existing approaches while substantially improving robustness under both seen and unseen speech reconstruction models.

Hybrid HE/2PC private CNN inference remains bottlenecked by prime-modulus homomorphic arithmetic in convolution and by a precision flow that runs ReLU at doubled bitwidth before invoking a separate truncation protocol. We present Jaguar, a system built on a single design choice–a power-of-two ciphertext ring–that addresses both. The choice enables SPA-Conv, a coefficient-domain convolution kernel that replaces NTT-centric polynomial multiplication with scalar-polynomial accumulation, and an exact ciphertext-side truncation by local right shifts that lets ReLU run directly at the target fixed-point precision and eliminates the post-ReLU truncation protocol. Where NTT remains genuinely useful–at the client, for the single polynomial multiplication during decryption–we recover it through an auxiliary NTT prime, preserving the power-of-two protocol substrate while keeping decryption O(N log N). On ImageNet-scale ResNet-18, ResNet-50, and MobileNetV2 with AVX disabled, Jaguar achieves 2.07-3.72x lower end-to-end latency than Cheetah and 2.16-3.36x lower than Rhombus, with 1.16-1.76x lower communication than Cheetah.

Large Language Models (LLMs) are increasingly used for code generation, raising concerns that they may be misused to produce malicious code. Meanwhile, Grammar-Constrained Decoding (GCD) has been widely adopted to improve the reliability of LLM-generated code by enforcing syntactic validity. In this paper, we reveal a counterintuitive risk: this reliability-oriented technique can itself become an attack surface. We uncover a new jailbreak attack, termed CodeSpear, that exploits GCD to induce LLMs into generating malicious code. Our experiments show that simply applying a benign code grammar constraint can effectively jailbreak LLMs. To address this vulnerability, we propose CodeShield, a safety alignment approach that robustly preserves safe behavior even under attacker-controlled grammar constraints. CodeShield aligns the model in the code modality by teaching it to generate honeypot code under GCD. Such code is semantically harmless, so it does not implement the malicious request, and structurally diverse, so it is difficult to suppress through grammar tightening. At the same time, CodeShield still preserves natural-language refusals when natural language is available. Experiments on 10 popular LLMs across 4 benchmarks show that CodeSpear outperforms representative jailbreak baselines and increases the attack success rate by more than 30 percentage points on average. CodeShield also restores safety under CodeSpear while preserving benign utility. Our findings reveal a fundamental risk of GCD and call for greater attention to its potential security implications.

Trustworthy AI requires reliable data-processing pipelines, not only robust downstream predictive models. As an upstream component, data summarization determines which information is retained and passed to subsequent learning or decision modules. Therefore, adversarial perturbations to the summarization process can compromise trustworthy AI in an upstream manner: they may alter the selected summary, reduce its representativeness, and further degrade the utility of subsequent learning tasks. In this paper, we study adversarial attacks on continuous data summarization under similarity-level perturbations through DR-submodular optimization. We show that a class of multi-resolution image summarization objectives can be formulated as multilinear extensions of non-negative submodular set functions and satisfy DR-submodularity with $m$-weak monotonicity. We then formulate multi-target attack generation as a min-max problem, where one admissible perturbation of the similarity structure is optimized to degrade multiple target summarization models. To mitigate such perturbations, we formulate robust defense against mixed attack types as a regularized max-min problem. For both problems, we develop approximation algorithms with theoretical guarantees. Experiments on real-data and controlled clustered benchmarks show that the proposed attack is effective in representative low-to-moderate budget regimes and can induce downstream task-performance loss. The proposed defense improves the robustness–mitigation trade-off in structured settings, while also revealing the parameter sensitivity of robust protection on real data.

The rapid growth of consumer IoT devices has introduced unprecedented challenges in trustworthy anomaly detection against AI-enabled cyber threats, requiring real-time, privacy-preserving, and scalable defense mechanisms. Traditional centralized strategies face critical limitations, including communication bottlenecks, single points of failure, and privacy vulnerabilities when processing distributed consumer data. We propose SwarmSense-DNN, a novel decentralized neural framework employing swarm intelligence for secure, cooperative anomaly detection across distributed IoT environments. The framework integrates autonomous agents with deep neural networks to form a self-organizing defense system that detects evolving anomalies without centralized coordination. It utilizes hierarchical federated learning with graph neural networks and attention mechanisms to capture local and global anomaly behaviors while ensuring data privacy. Extensive experiments demonstrate SwarmSense-DNN's superior performance: it achieves 95.44% average detection accuracy across five benchmark datasets while reducing communication overhead by 67%. The framework maintains robust resilience against adversarial threats through differential privacy safeguards and demonstrates strong fault tolerance under node failures and AI-enabled attacks.

We consider the problem of privately releasing a $k$-dimensional vector under updates: Starting with a zero vector, at times $t_1, t_2,\dots$ the vector is updated by adding $x^{(1)}, x^{(2)},\dots$, respectively. For positive integers $T$, $k$ we model the updates as a data set $\{(t_i, x^{(i)})\}_i$, where $t_i \in [T]$ and $x^{(i)} \in B_k$ (the $k$-dimensional unit ball). Two such data sets are said to be neighboring if their symmetric difference has size at most $1$. The continual release consists of the sum $A^{(t)} = \sum_{i \; : \; t_i \leq t} x^{(i)}$ for each time step $t=1,\dots,T$. Classical continual release techniques allow us to release an approximation of $A^{(1)},\dots,A^{(T)}$ with additive noise of magnitude $\text{polylog}(T)$, computed in time $O(kT)$, even in the on-line, adaptive case where data is continually revealed for the current time step. Motivated by private sketching techniques, we consider the setting where only a subset of entries in $A^{(t)}$ need to be released at time step $t$. Our new result is that it is possible to sample any desired entry in a given noise vector in constant time while reproducing exactly the distribution of the binary tree mechanism with Gaussian noise. The improvement on the known time bound of $O(\log T)$ comes from a new data structure that allows us to sample a new noise value with the correct correlations in constant time using Brownian bridges. We present two data management applications, of independent interest, that use our technique in conjunction with differentially private CountSketches: 1) A dynamic data structure for orthogonal range counting queries with a better privacy/accuracy/space trade-off than previous data structures, and 2) Join size estimation, where in addition we show improved high-probability bounds.

State root computation dominates (78%) blockchain block processing time. Ethereum's canonical authenticated data structure, i.e., Merkle Patricia Trie (MPT), suffers from severe tree-height growth and is vulnerable to Nurgle attacks (SP'24), where adversaries inflate path depth via hash collisions and degrade system performance at negligible cost. Existing defenses increase node fanout (span) to bound tree height, but higher span inflates proof size exponentially. Prior work mitigates this trade-off using vector commitments, at the cost of trusted setup or expensive verification. We present Mhot, a height-optimal authenticated data structure for blockchain state commitment that preserves standard hash-based verification without trusted setup. Unlike MPT's fixed-prefix indexing, which couples span and fanout exponentially, Mhot indexes by discriminative bits that actually distinguish keys, achieving adaptive span with linear fanout coupling and provably minimal height. To prevent high fanout from inflating proofs, we introduce hierarchical proofs, a two-layer Merkle construction that reduces per-node proof overhead from O(k) to O(log k). On Ethereum mainnet workloads, Mhot achieves up to 9X higher write throughput, 4X lower write amplification, and 2X smaller proofs than MPT. Under Nurgle attacks, even when the adversary consumes an entire block's gas budget, Mhot maintains a 0% attack success rate (v.s., 99.97% for MPT). Our results, somewhat surprisingly, show that height optimality (not new crypto primitives!) is the key abstraction for scalable and attack-resilient blockchain state commitment.

Industry has embraced Zero Trust (ZT) architectural tenets and implementations for cloud-native environments, following stricter security requirements to both internal and external tenants. Among others, these approaches combine fine-grained identity management and monitoring for both inventorying and better analysing the devices' security posture for overall protection, along with strict separation of concerns and isolation to enforce minimal privilege. Networking-wise, ZT approaches rely as well on isolation and least privilege; enacted by separate, secure tunnels per tenant connecting to a given infrastructure. Such implementations can also be applied to the connectivity within and towards experimental infrastructures. In this sense, this work contributes the design and evaluation of a cloud-native VPN-as-a-Service (VPNaaS) that can be (i) easily orchestrated to deploy on-the-fly, separate tunnels per each tenant remotely connecting to the infrastructure; (ii) integrated with common Identity and Access Management (IAM) tools, key to ZT deployments; and (iii) adapt to computing- or entropy- constrained environments. This solution is customisable and allows, among others, to select from RSA or Elliptic Curves (EC) as key generation algorithm and their parameters to achieve more secure keys and adapt to resource-constrained environments.