Papers

Modern phishing campaigns increasingly evade snapshot-based URL classifiers using interaction gates (e.g., checkbox/slider challenges), delayed content rendering, and logo-less credential harvesters. This shifts URL triage from static classification toward an interactive forensics task: an analyst must actively navigate the page while isolating themselves from potential runtime exploits. We present TraceScope, a decoupled triage pipeline that operationalizes this workflow at scale. To prevent the observer effect and ensure safety, a sandboxed operator agent drives a real GUI browser guided by visual motivation to elicit page behavior, freezing the session into an immutable evidence bundle. Separately, an adjudicator agent circumvents LLM context limitations by querying evidence on demand to verify a MITRE ATT&CK checklist, and generates an audit-ready report with extracted indicators of compromise (IOCs) and a final verdict. Evaluated on 708 reachable URLs from existing dataset (241 verified phishing from PhishTank and 467 benign from Tranco-derived crawling), TraceScope achieves 0.94 precision and 0.78 recall, substantially improving recall over three prior visual/reference-based classifiers while producing reproducible, analyst-grade evidence suitable for review. More importantly, we manually curated a dataset of real-world phishing emails to evaluate our system in a practical setting. Our evaluation reveals that TraceScope demonstrates superior performance in a real-world scenario as well, successfully detecting sophisticated phishing attempts that current state-of-the-art defenses fail to identify.

Large language model (LLM) agents increasingly rely on skills to package reusable capabilities through instructions, tools, and resources. High-quality skills embed expert knowledge, curated workflows, and execution constraints into agents, fueling a growing skill economy through their value and scalability. Yet this ecosystem also creates a new attack surface, as adversaries can interact with public agent interfaces to extract hidden proprietary skill content. We present the first systematic study of black-box skill stealing against LLM agent systems. Compared with conventional system prompt stealing, skill stealing targets modular and structured capability packages whose leakage is directly actionable for copying, redistribution, and monetization, making the resulting harm potentially greater. To study this threat, we derive an attack taxonomy from prior prompt-stealing methods and build an automated stealing prompt generation agent. Starting from model-generated seed prompts, the framework expands attacks through scenario rationalization and structure injection while enforcing diversity via embedding-based filtering, yielding a reproducible pipeline for evaluating proprietary agent systems. We evaluate these attacks across commercial agent platforms and representative LLMs. Our results show that agent skills can often be extracted easily, posing a serious copyright risk. To mitigate this threat, we design defenses across the agent pipeline, focusing on input, inference, and output phase. Although these defenses substantially reduce leakage, the attack remains inexpensive and repeatable, and a single successful attempt is sufficient to compromise the protected skill. Overall, our findings suggest that these copyright risks remain largely overlooked across proprietary agent ecosystems, motivating stronger protection mechanisms.

Near-field millimeter-wave (mmWave) imaging is widely deployed in safety-critical applications such as airport passenger screening, yet its own security remains largely unexplored. This paper presents a systematic study of the adversarial robustness of mmWave imaging algorithms under waveform-domain physical attacks that directly manipulate the image reconstruction process. We propose a practical white-box adversarial model and develop a differential imaging attack framework that leverages the differentiable imaging pipeline to optimize attack waveforms. We also construct a real measured dataset of clean and attack waveforms using a mmWave imaging testbed. Experiments on 10 representative imaging algorithms show that mmWave imaging is highly vulnerable to such attacks, enabling an adversary to conceal or alter targets with moderate transmission power. Surprisingly, deep-learning-based imaging algorithms demonstrate higher robustness than classical algorithms. These findings expose critical security risks and motivate the development of robust and secure mmWave imaging systems.

The growing application of large language models (LLMs) in safety-critical domains has raised urgent concerns about their security. Many recent studies have demonstrated the feasibility of backdoor attacks against LLMs. However, existing methods suffer from three key shortcomings: explicit trigger patterns that compromise naturalness, unreliable injection of attacker-specified payloads in long-form generation, and incompletely specified threat models that obscure how backdoors are delivered and activated in practice. To address these gaps, we present BadStyle, a complete backdoor attack framework and pipeline. BadStyle leverages an LLM as a poisoned sample generator to construct natural and stealthy poisoned samples that carry imperceptible style-level triggers while preserving semantics and fluency. To stabilize payload injection during fine-tuning, we design an auxiliary target loss that reinforces the attacker-specified target content in responses to poisoned inputs and penalizes its emergence in benign responses. We further ground the attack in a realistic threat model and systematically evaluate BadStyle under both prompt-induced and PEFT-based injection strategies. Extensive experiments across seven victim LLMs, including LLaMA, Phi, DeepSeek, and GPT series, demonstrate that BadStyle achieves high attack success rates (ASRs) while maintaining strong stealthiness. The proposed auxiliary target loss substantially improves the stability of backdoor activation, yielding an average ASR improvement of around 30% across style-level triggers. Even in downstream deployment scenarios unknown during injection, the implanted backdoor remains effective. Moreover, BadStyle consistently evades representative input-level defenses and bypasses output-level defenses through simple camouflage.

Technology for security operations centers (SOCs) has a storied history of slow adoption due to concerns about trust and reliability. These concerns are amplified with artificial intelligence, particularly large language models (LLMs), which exhibit issues such as hallucinations and inconsistent outputs. To assess whether LLM-based tools can improve SOC efficiency, we embedded two PhD researchers within a multinational company SOC for six months of ethnographic fieldwork. We identified recurring challenges, such as repetitive tasks, fragmented/unclear data, and tooling bottlenecks, and collaborated directly with practitioners to develop LLM companion tools aligned with their operational needs. Iterative refinement reduced workflow disruption and improved interpretability, leading from skepticism to sustained adoption. Ethnographic analysis indicates that this shift was enabled by our sociotechnical co-creation process consistent with Nonaka's SECI model. This framework explains the common challenges in traditional SOC technology adoption, including workflow misalignment, rigidity against evolving threats and internal requirements, and stagnation over time. Our findings show that the co-creation approach can overcome these old barriers and create a new paradigm for creating usable technology for cybersecurity operations.

Past attacks against industrial control systems (ICS) show that adversaries often target both the ICS network and the physical process to achieve potential catastrophic impact. To secure ICS, intrusion detection systems promise timely uncovering of such adversaries. However, as these detection mechanisms typically focus on isolated characteristics of ICS (e.g., packet timings), multiple detection systems have to be deployed in parallel, complicating their operation in practice. In this work, to spur discussion and further research, we present challenges encountered during our research towards a holistic intrusion detection system aiming to cover all dimensions of an ICS.

The proliferation of Internet of Things (IoT) devices has significantly expanded attack surfaces, making IoT ecosystems particularly susceptible to sophisticated cyber threats. To address this challenge, this work introduces A-THENA, a lightweight early intrusion detection system (EIDS) that significantly extends preliminary findings on time-aware encodings. A-THENA employs an advanced Transformer-based architecture augmented with a generalized Time-Aware Hybrid Encoding (THE), integrating packet timestamps to effectively capture temporal dynamics essential for accurate and early threat detection. The proposed system further employs a Network-Specific Augmentation (NA) pipeline, which enhances model robustness and generalization. We evaluate A-THENA on three benchmark IoT intrusion detection datasets-CICIoT23-WEB, MQTT-IoT-IDS2020, and IoTID20-where it consistently achieves strong performance. Averaged across all three datasets, it improves accuracy by 6.88 percentage points over the best-performing traditional positional encoding, 3.69 points over the strongest feature-based model, 6.17 points over the leading time-aware alternatives, and 5.11 points over related models, while achieving near-zero false alarms and false negatives. To assess real-world feasibility, we deploy A-THENA on the Raspberry Pi Zero 2 W, demonstrating its ability to perform real-time intrusion detection with minimal latency and memory usage. These results establish A-THENA as an agile, practical, and highly effective solution for securing IoT networks.

The automotive domain is transitioning: vehicles act as rolling servers, persistently connected to numerous external entities. This connectivity, combined with rising on-board computing power for advanced driver assistance systems and similar use cases, creates escalating challenges for securing automotive network architectures. This work advances the security analysis of internet-connected automotive network architectures and their protocols. We introduce a strong, active adversary model tailored to the automotive domain. We substantially extend security protocol verification possible based on Attack Resilience Hyperproperties (ARHs) by introducing a verification-orchestration algorithm. Furthermore, we provide methods for comparative attribution of security property invalidations to specific, ne-grained component compromises. We present a novel integration of formal verification and process mining. By utilizing ARH counterexample traces for process mining, we systematically identify and aggregate attacker behavior that causes security property invalidations. This pipeline enables in-depth understanding of root causes and attack paths leading to protocol-security invalidations. We demonstrate real-world applicability through a prototype and case study on the secure transmission of battery management system data within an automotive network architecture.

Contemporary cybersecurity governance assumes that professionals apply risk reasoning. Yet major organisational failures persist despite investment in tools, staffing, and credentials. This study investigates the structural source of that paradox. Cybersecurity speaks the language of risk, but its training architecture has shaped the profession to think in terms of threats. A sequential mixed-methods design integrated four analyses; NLP of the NIST NICE Framework v2.0.0 (2,111 TKS statements), SEM (n = 126 cybersecurity professionals), a control-group comparison (n = 133 general professionals), and thematic coding of seven leadership interviews. Four convergent findings emerged. First, "likelihood" and "probability" appear zero times across all TKS statements. Risk management content accounts for 4.5% of high-confidence semantic classifications, ranking 18th of 29 competency domains. NICE codifies threat-management activity while invoking risk mainly at the category level. Second, SEM showed that training exposure significantly predicts risk management competence directly and indirectly through conceptual salience, for a total effect of Beta = .629. However, the theoretically four-dimensional competence construct collapsed into a single factor, indicating epistemic compression. Third, cybersecurity professionals showed no measurable advantage over the general professional population in foundational risk reasoning; only 11.9% showed high differentiation. Fourth, all seven leaders expected Likelihood x Impact reasoning, yet five did not articulate the formula themselves. These findings support a structural conclusion: cybersecurity has taken professional form as a threat-management discipline that has borrowed risk vocabulary. Remediation requires redesign of professional formation, not marginal curriculum reform.

Differential privacy (DP) is a mathematical framework that guarantees individual privacy; however, systematic evaluation of its impact on statistical utility in survival analyses remains limited. In this study, we systematically evaluated the impact of DP mechanisms (Laplace mechanism and Randomized Response) with data-driven clipping bounds on the Cox proportional hazards model, using 5 clinical datasets ($n = 168$–$6{,}524$), 15 levels of $\varepsilon$ (0.1–1000), and $B = 1{,}000$ Monte Carlo iterations. The data-driven clipping bounds used here are observed min/max and therefore do not provide formal $\varepsilon$-DP guarantees; the results represent an optimistic lower bound on utility degradation under formal DP. We compared three types of input perturbations (covariates only, all inputs, and the discrete-time model) with output perturbations (dfbeta-based sensitivity), using loss of significance rate (LSR), C-index, and coefficient bias as metrics. At standard DP levels ($\varepsilon \leq 1$), approximately 90% (90–94%) of the significant covariates lost significance, even in the largest dataset ($n = 6{,}524$), and the predictive performance approached random levels (test C-index $\approx 0.5$) under many conditions. Among the input perturbation approaches, perturbing only covariates preserved the risk-set structure and achieved the best recovery, whereas output perturbation (dfbeta-based sensitivity) maintained near-baseline performance at $\varepsilon \geq 5$. At $n \approx 3{,}000$, the significance recovered rapidly at $\varepsilon = 3$–10; however, in practice, $\varepsilon \geq 10$ (for predictive performance) to $\varepsilon \geq 30$–60 (for significance preservation) is required. In the moderate-to-high $\varepsilon$ range, false-positive rates increased for variables whose baseline $p$-values were near the significance threshold.

Model Context Protocol (MCP) enables tool-integrated LLM agents, but its third-party tool-server ecosystem expands software supply-chain risk across tool metadata, untrusted outputs, cross-tool flows, multi-modal inputs, and privileged sink actions. Existing MCP benchmarks mainly measure robustness to malicious inputs, offering limited support for dependency vetting, semantic metadata auditing, and hardening regression. We present MCP Pitfall Lab, a protocol-aware security testing framework that models developer pitfalls as reproducible scenarios and validates outcomes using MCP traces and objective validators rather than agent self-report. We also introduce Semantic MCP-Bill-of-Material(MCP-BOM), representing that augments component inventory with security-relevant tool semantics, including descriptions, schemas, high-risk parameters, source/sink roles, trust boundaries, policy hooks, audit support, and trace provenance. Across 2,579 validator-completed runs over four models, Pitfall Lab observes 31.9% overall attack success rate(ASR), with multi-modal injection strongest at 38.7%. Semantic static fields detect pitfalls involving policy-bearing tool descriptions, permissive schemas, missing audit support, and absent server-side validation with F1=0.727, while BOM-backed findings drop from 27 to 16 after hardening. Decomposed risk metrics show Control Coverage increasing from 0.173 to 0.697 and Residual Risk decreasing from 15.31 to 6.09. A trace-linked case study further demonstrates that cross-tool forwarding and image-to-tool leakage require runtime provenance rather than static manifest fields alone. Overall, Pitfall Lab reframes MCP tool servers as AI software-supply-chain dependencies and provides BOM-backed artifacts for review, hardening, and trace-grounded regression testing

Similar to a strategic interaction between rational and intelligent agents, cryptography problems can be examined through the prism of game theory. In this setting, the agent aiming to protect a message is called the defender, while the one attempting to decrypt it, generally for malicious purposes, is the attacker. To strengthen security in cryptography, various strategies have been developed, among which hybridization stands out as a key concept in modern cryptographic design. This strategy allows the defender to select among different encryption algorithms (classical, post-quantum, or hybrid) while carefully balancing security and operational costs. On the other side, the attacker, limited by available resources, chooses cryptanalysis methods capable of breaching the selected algorithm. We model this interaction as a Stackelberg cryptographic hybridization problem under resource constraints. Here, the defender randomizes over encryption algorithms, and the attacker observes the choice before selecting suitable cryptanalysis methods. The attacker's decision is framed as a conditional optimization problem, which we refer to as the “attacker subgame”. We then propose a dynamic programming approach for the attacker's subgame, while the defender's Stackelberg optimization is formulated as a linear program.

Protecting patient privacy in clinical narratives is essential for enabling secondary use of healthcare data under regulations such as GDPR and HIPAA. While manual de-identification remains the gold standard, it is costly and slow, motivating the need for automated methods that combine privacy guarantees with high utility. Most automated text de-identification pipelines employed named entity recognition (NER) to identify protected entities for redaction. Although methods based on differential privacy (DP) provide formal privacy guarantees, more recently also large language models (LLMs) are increasingly used for text de-identification in the clinical domain. In this work, we present the first comparative study of DP, NER, and LLMs for Dutch clinical text de-identification. We investigate these methods separately as well as hybrid strategies that apply NER or LLM preprocessing prior to DP, and assess performance in terms of privacy leakage and extrinsic evaluation (entity and relation classification). We show that DP mechanisms alone degrade utility substantially, but combining them with linguistic preprocessing, especially LLM-based redaction, significantly improves the privacy-utility trade-off.

Poisoning-based backdoor attacks pose significant threats to deep neural networks by embedding triggers in training data, causing models to misclassify triggered inputs as adversary-specified labels while maintaining performance on clean data. Existing poison restraint-based defenses often suffer from inadequate detection against specific attack variants and compromise model utility through unlearning methods that lead to accuracy degradation. This paper conducts a comprehensive analysis of backdoor attack dynamics during model training, revealing that poisoned samples form isolated clusters in latent space early on, with triggers acting as dominant features distinct from benign ones. Leveraging these insights, we propose Cluster Segregation Concealment (CSC), a novel poison suppression defense. CSC first trains a deep neural network via standard supervised learning while segregating poisoned samples through feature extraction from early epochs, DBSCAN clustering, and identification of anomalous clusters based on class diversity and density metrics. In the concealment stage, identified poisoned samples are relabeled to a virtual class, and the model's classifier is fine-tuned using cross-entropy loss to replace the backdoor association with a benign virtual linkage, preserving overall accuracy. CSC was evaluated on four benchmark datasets against twelve poisoning-based attacks, CSC outperforms nine state-of-the-art defenses by reducing average attack success rates to near zero with minimal clean accuracy loss. Contributions include robust backdoor patterns identification, an effective concealment mechanism, and superior empirical validation, advancing trustworthy artificial intelligence.

Steganography embeds secret messages in seemingly innocuous carriers for covert communication under surveillance. Current Provably Secure Steganography (PSS) schemes based on language models can guarantee computational indistinguishability between the covertext and stegotext. However, achieving high embedding capacity remains a challenge for existing PSS. The inefficient entropy utilization renders them not well-suited for Large Language Models (LLMs), whose inherent low-entropy tendencies severely constrain feasible embedding capacity. To address this, we propose a provably secure steganography scheme with a theoretically proved high capacity. Our scheme is based on the concept of list decoding: it maintains a set of candidates that contain the correct secret message, instead of directly finding the correct message with more effort. This strategy fully utilizes the information content of the generated text, yielding higher capacity. To ensure the correctness of our scheme, we further introduce a suffix-matching mechanism to distinguish the correct secret message from the candidates. We provide theoretical proofs for both the security and correctness of our scheme, alongside a derivation of its theoretical capacity lower bound. Our approach is plug-and-play, requiring only a direct replacement of the model's standard random sampling module. Experiments on three LLMs and seven PSS baselines demonstrate that our method achieves computational efficiency comparable to prior PSS schemes while delivering a substantial improvement in embedding capacity.

Deep learning has emerged as a powerful approach for malware detection, demonstrating impressive accuracy across various data representations. However, these models face critical limitations in real-world, non-stationary environments where both malware characteristics and detection systems continuously evolve. Our research investigates a fundamental security question: Can an attacker generate adversarial malware samples that simultaneously evade classification and remain inconspicuous to drift monitoring mechanisms? We propose a novel approach that generates targeted adversarial examples in the classifier's standardized feature space, augmented with sophisticated similarity regularizers. By carefully constraining perturbations to maintain distributional similarity with clean malware, we create an optimization objective that balances targeted misclassification with drift signal minimization. We quantify the effectiveness of this approach by comprehensively comparing classifier output probabilities using multiple drift metrics. Our experiments demonstrate that similarity constraints can reduce output drift signals, with $\ell_2$ regularization showing the most promising results. We observe that perturbation budget significantly influences the evasion-detectability trade-off, with increased budget leading to higher attack success rates and more substantial drift indicators.

Enterprise LLM agents can dramatically improve workplace productivity, but their core capability, retrieving and using internal context to act on a user's behalf, also creates new risks for sensitive information leakage. We introduce CI-Work, a Contextual Integrity (CI)-grounded benchmark that simulates enterprise workflows across five information-flow directions and evaluates whether agents can convey essential content while withholding sensitive context in dense retrieval settings. Our evaluation of frontier models reveals that privacy failures are prevalent (violation rates range from 15.8%-50.9%, with leakage reaching up to 26.7%) and uncovers a counterintuitive trade-off critical for industrial deployment: higher task utility often correlates with increased privacy violations. Moreover, the massive scale of enterprise data and potential user behavior further amplify this vulnerability. Simply increasing model size or reasoning depth fails to address the problem. We conclude that safeguarding enterprise workflows requires a paradigm shift, moving beyond model-centric scaling toward context-centric architectures.

Automated code vulnerability detection is critical for software security, yet existing approaches face a fundamental trade-off between detection accuracy and computational cost. We propose a heterogeneous multi-agent architecture inspired by game-theoretic principles, combining cloud-based LLM experts with a local lightweight verifier. Our "3+1" architecture deploys three cloud-based expert agents (DeepSeek-V3) that analyze code from complementary perspectives - code structure, security patterns, and debugging logic - in parallel, while a local verifier (Qwen3-8B) performs adversarial validation at zero marginal cost. We formalize this design through a two-layer game framework: (1) a cooperative game among experts capturing super-additive value from diverse perspectives, and (2) an adversarial verification game modeling quality assurance incentives. Experiments on 262 real samples from the NIST Juliet Test Suite across 14 CWE types, with balanced vulnerable and benign classes, demonstrate that our approach achieves a 77.2% F1 score with 62.9% precision and 100% recall at $0.002 per sample - outperforming both a single-expert LLM baseline (F1 71.4%) and Cppcheck static analysis (MCC 0). The adversarial verifier significantly improves precision (+10.3 percentage points, p < 1e-6, McNemar's test) by filtering false positives, while parallel execution achieves a 3.0x speedup. Our work demonstrates that game-theoretic design principles can guide effective heterogeneous multi-agent architectures for cost-sensitive software engineering tasks.

This paper presents ECCFROG522PP, a 522-bit prime-field elliptic curve in short Weierstrass form, designed with a focus on deterministic generation and public reproducibility. The central design principle is that all critical parameters are derived from a fixed public seed through a transparent and verifiable procedure. While many deployed systems rely on NIST P-256 and secp256k1, which target approximately 128-bit classical security, higher security applications typically consider curves such as NIST P-521, Curve448, and Brainpool P512. ECCFROG522PP is intended for the same general classical security range as P-521, with emphasis on transparency, auditability, and reproducibility rather than performance optimization. The curve parameters are generated through a BLAKE3-based deterministic pipeline with publicly specified indices. The resulting construction has prime order, cofactor one, and a deterministically derived base point of full order. The quadratic twist has a large proven prime factor, and the construction includes a documented lower bound on the embedding degree together with standard sanity checks against low embedding degree reductions and basic CM discriminant anomalies. The full generation and validation procedure can be reproduced end to end from public artifacts and reference scripts, enabling independent verification of all parameters and checks.

The rapid integration of artificial intelligence (AI) into Internet of Things (IoT) and edge computing systems has intensified the need for robust, hardware-rooted trust mechanisms capable of ensuring device authenticity and AI model integrity under strict resource and security constraints. This survey reviews and synthesizes existing literature on hardware-rooted trust mechanisms for AI-enabled IoT systems. It systematically examines and compares representative trust anchor mechanisms, including Trusted Platform Module (TPM)-based measurement and attestation, silicon and FPGA-based Physical Unclonable Functions (PUFs), hybrid container-aware hardware roots of trust, and software-only security approaches. The analysis highlights how hardware-rooted solutions generally provide stronger protection against physical tampering and device cloning compared to software-only approaches, particularly in adversarial and physically exposed environments, while hybrid designs extend hardware trust into runtime and containerized environments commonly used in modern edge deployments. By evaluating trade-offs among security strength, scalability, cost, and deployment complexity, the study shows that PUF-based and hybrid trust anchors offer a promising balance for large-scale, AI-enabled IoT systems, whereas software-only trust mechanisms remain insufficient in adversarial and physically exposed settings. The presented comparison aims to clarify current design challenges and guide future development of trustworthy AI-enabled IoT platforms.

Transaction simulation is an important subsystem of block building, denial of whose service could lead to severe damage to the blockchain ecosystem and transaction delivery. Denial of block building has been studied, where the existing attack designs either target single-round builders, such as ConditionalExhaust (USENIX Security '24), or target two-round builders, by exploiting cross-round inconsistency, such as GhostTX (USENIX Security '24) and denial of sequencers (CCS '25). This work examines the denial-of-service security of multi-round transaction simulation under a new exploit: inter-transaction dependency that manifests in smart-contract state.

Many approaches to LLM red-teaming leverage an attacker LLM to discover jailbreaks against a target. Several of them task the attacker with identifying effective strategies through trial and error, resulting in a semantically limited range of successes. Another approach discovers diverse attacks by combining crowdsourced harmful queries and tactics into instructions for the attacker, but does so at random, limiting effectiveness. This article introduces a novel framework, Adaptive Instruction Composition, that combines crowdsourced texts according to an adaptive mechanism trained to jointly optimize effectiveness with diversity. We use reinforcement learning to balance exploration with exploitation in a combinatorial space of instructions to guide the attacker toward diverse generations tailored to target vulnerabilities. We demonstrate that our approach substantially outperforms random combination on a set of effectiveness and diversity metrics, even under model transfer. Further, we show that it surpasses a host of recent adaptive approaches on Harmbench. We employ a lightweight neural contextual bandit that adapts to contrastive embedding inputs, and provide ablations suggesting that the contrastive pretraining enables the network to rapidly generalize and scale to the massive space as it learns.

This paper studies 43-class malware type classification on MalNet-Image Tiny, a public benchmark derived from Android APK files. The goal is to assess whether a compact image classifier benefits from four components evaluated in a controlled ablation: a feature pyramid network (FPN) for scale variation induced by resizing binaries of different lengths, ImageNet pretraining, lightweight augmentation through Mixup and TrivialAugment, and schedule-free AdamW optimization. All experiments use a ResNet18 backbone and the provided train/validation/test split. Reproducing the benchmark-style configuration yields macro-F1 (F1_macro) of 0.6510, consistent with the reported baseline of approximately 0.65. Replacing the optimizer with schedule-free AdamW and using unweighted cross-entropy increases F1_macro to 0.6535 in 10 epochs, compared with 96 epochs for the reproduced baseline. The best configuration combines pretraining, Mixup, TrivialAugment, and FPN, reaching F1_macro=0.6927, P_macro=0.7707, AUC_macro=0.9556, and L_test=0.8536. The ablation indicates that the largest gains in F1_macro arise from pretraining and augmentation, whereas FPN mainly improves P_macro, AUC_macro, and L_test in the strongest configuration.

AI-agent guardrails are memoryless: each message is judged in isolation, so an adversary who spreads a single attack across dozens of sessions slips past every session-bound detector because only the aggregate carries the payload. We make three contributions to cross-session threat detection. (1) Dataset. CSTM-Bench is 26 executable attack taxonomies classified by kill-chain stage and cross-session operation (accumulate, compose, launder, inject_on_reader), each bound to one of seven identity anchors that ground-truth "violation" as a policy predicate, plus matched Benign-pristine and Benign-hard confounders. Released on Hugging Face as intrinsec-ai/cstm-bench with two 54-scenario splits: dilution (compositional) and cross_session (12 isolation-invisible scenarios produced by a closed-loop rewriter that softens surface phrasing while preserving cross-session artefacts). (2) Measurement. Framing cross-session detection as an information bottleneck to a downstream correlator LLM, we find that a session-bound judge and a Full-Log Correlator concatenating every prompt into one long-context call both lose roughly half their attack recall moving from dilution to cross_session, well inside any frontier context window. Scope: 54 scenarios per shard, one correlator family (Anthropic Claude), no prompt optimisation; we release it to motivate larger, multi-provider datasets. (3) Algorithm and metric. A bounded-memory Coreset Memory Reader retaining highest-signal fragments at $K=50$ is the only reader whose recall survives both shards. Because ranker reshuffles break KV-cache prefix reuse, we promote $\mathrm{CSR\_prefix}$ (ordered prefix stability, LLM-free) to a first-class metric and fuse it with detection into $\mathrm{CSTM} = 0.7 F_1(\mathrm{CSDA@action}, \mathrm{precision}) + 0.3 \mathrm{CSR\_prefix}$, benchmarking rankers on a single Pareto of recall versus serving stability.

Automated vulnerability detection tools are widely used to identify security vulnerabilities in software dependencies. However, the evaluation of such tools remains challenging due to the heterogeneous structure of vulnerability data sources, inconsistent identifier schemes, and ambiguities in version range specifications. In this paper, we present an empirical evaluation of vulnerability detection across multiple software ecosystems using a curated ground-truth dataset derived from the Open Source Vulnerabilities (OSV) database. The dataset explicitly maps vulnerabilities to concrete package versions and enables a systematic comparison of detection results across different tools and services. Since vulnerability databases such as OSV are continuously updated, the dataset used in this study represents a snapshot of the vulnerability landscape at the time of the evaluation. To support reproducibility and future studies, we provide an open-source tool that automatically reconstructs the dataset from the current OSV database using the methodology described in this paper. Our evaluation highlights systematic differences between vulnerability detection systems and demonstrates the importance of transparent dataset construction for reproducible empirical security research.

Third-party Large Language Model (LLM) API gateways are rapidly emerging as unified access points to models offered by multiple vendors. However, the internal routing, caching, and billing policies of these gateways are largely undisclosed, leaving users with limited visibility into whether requests are served by the advertised models, whether responses remain faithful to upstream APIs, or whether invoices accurately reflect public pricing policies. To address this gap, we introduce GateScope, a lightweight black-box measurement framework for evaluating behavioral consistency and operational transparency in commercial LLM gateways. GateScope is designed to detect key misbehaviors, including model downgrading or switching, silent truncation, billing inaccuracies, and instability in latency by auditing gateways along four critical dimensions: response content analysis, multi-turn conversation performance, billing accuracy, and latency characteristics. Our measurements across 10 real-world commercial LLM API gateways reveal frequent gaps between expected and actual behaviors, including silent model substitutions, degraded memory retention, deviations from announced pricing, and substantial variation in latency stability across platforms.

Layer-2 (L2) protocols address the fundamental limitations of Layer-1 (L1) blockchains by offloading computation while anchoring trust to the parent chain. This architectural shift, while boosting throughput, introduces a new, complex security surface defined by off-chain components like sequencers, bridges, and data availability mechanisms. Prior literature[31][33] offers fragmented views of this risk. This paper presents the first unified, security-focused survey that rigorously maps L2 architecture to its underlying cryptographic security. We dissect the technical progression from L1 primitives to the core of modern L2s, analyzing the security assumptions(Discrete Logarithm, Computational Diffie-Hellman, Bilinear Diffie-Hellman) of ZK frameworks (Groth16, Plonk) and their corresponding commitment schemes (KZG, IPA). We formalize a comprehensive L2 threat model encompassing sequencer liveness, bridge exploits, and data-availability failures. This work serves as an accessible yet rigorous reference for researchers and developers to reason about L2 security from a deep crypto-mathematical perspective.

Software security assurance relies on effective vulnerability detection and patching, yet determining whether a patch fully eliminates risk remains an underexplored challenge. Existing vulnerability benchmarks often treat patched functions as inherently benign, overlooking the possibility of residual security risks. In this work, we analyze vulnerable-benign C/C++ function pairs from PrimeVul, a benchmark dataset of real-world vulnerabilities. We use multiple code language models (Code LMs) to capture semantic similarity, complemented by Tree-sitter-based abstract syntax tree (AST) analysis to measure structural similarity. Building on these measures, we propose Residual Risk Scoring (RRS), a unified framework that integrates embedding-based semantic similarity, localized AST-based structural similarity, and cross-model agreement to estimate residual risk in code. Our analysis shows that benign functions often remain highly similar to their vulnerable counterparts both semantically and structurally, indicating potential persistence of residual risk. We further find that approximately $61\%$ of high-RRS C/C++ code pairs exhibit $13$ distinct categories of residual issues (e.g., null pointer dereferences, unsafe memory allocation), validated using state-of-the-art static analysis tools including Cppcheck, Clang-Tidy, and Facebook-Infer. These results demonstrate that code-level similarity provides a practical signal for prioritizing post-patch inspection and residual risk assessment. By providing a quantitative measure of residual risk, RRS supports software security assurance through risk-informed post-patch prioritization and patch validation.

Password-based authentication is one of the most commonly used methods for verifying user identities, and its widespread usage continues in virtual reality (VR) applications. As a result, various forms of attacks on password-based authentication in traditional environments such as keystroke inference and shoulder surfing, are still effective in VR applications. While keystroke inference attacks on virtual keyboards have been studied extensively, few efforts have developed an effective and cost-efficient defense strategy to mitigate keystroke inferences in VR. To address this gap, this paper presents a novel QWERTY keyboard called VRSafe that is resilient to keystroke inference attacks. The proposed keyboard carefully introduces false positive keystrokes into the information collected by attackers during the typing process, making the inference of the original password difficult. VRSafe also incorporates a novel malicious login detector that can effectively identify unauthorized login attempts using credentials inferred from keystroke inference attacks with high detection rate and minimal time and memory cost. The proposed design is evaluated through both simulation experiments and a real-world user study, and the results show that VRSafe can significantly reduce the accuracy of keystroke inference attacks while incurring a modest overhead from a usability standpoint.

The growth of agentic AI has drawn significant attention to function calling Large Language Models (LLMs), which are designed to extend the capabilities of AI-powered system by invoking external functions. Injection and jailbreaking attacks have been extensively explored to showcase the vulnerabilities of LLMs to user prompt manipulation. The expanded capabilities of agentic models introduce further vulnerabilities via their function calling interface. Recent work in LLM security showed that function calling can be abused, leading to data tampering and theft, causing disruptive behavior such as endless loops, or causing LLMs to produce harmful content in the style of jailbreaking attacks. This paper introduces a novel function hijacking attack (FHA) that manipulates the tool selection process of agentic models to force the invocation of an attacker-chosen function. While existing attacks focus on semantic preference of the model for function-calling tasks, we show that FHA is largely agnostic to the context semantics and remains effective across domains and function sets. We demonstrate that FHA generalizes to unseen queries and payload perturbations under a fixed target model, reaching 62.5% to 81.9% ASR on held-out queries across 4 function-calling LLMs (instructed and reasoning models), evaluated on the Berkeley Function Calling Leaderboard (BFCL). We further evaluate the cross-model transferability of FHA, showing that FHA can be transferred to other model sizes and families (11.2-27.6% ASR). Our findings further demonstrate the need for strong guardrails and modules for agentic systems.