cybersecurity teams

Abstract: Under tight feature and compute budgets, classical threat detection pipelines often degrade on near-decision-boundary events. Small quantum processors are now available, but existing work inadequately shows whether quantum components improve end-to-end threat detection under the above resource constrained conditions. This paper tries to address this gap with a hybrid architecture that uses a compact multilayer perceptron layer to compress the information in data and then routes the processed features to a few qubit quantum heads implemented in quantum support vector machine (QSVM) and variational quantum circuit (VQC) models. On a simulation platform, we benchmark these hybrid models against classical models with comparable parameter budgets on two representative cybersecurity tasks, network intrusion detection on NSL-KDD dataset and spam filtering on Ling-Spam dataset. To validate their precision on real quantum hardware, we deploy the best 4-qubit QSVM model on an IBM Quantum device with noise-aware execution, evaluated on a smaller sub-dataset. In the results, shallow quantum heads consistently match, and on difficult near-boundary cases modestly reduce missed attacks and false alarms compared to classical models using the same features. Hardware validation results track the simulation behavior closely enough that the remaining gap is dominated by device noise rather than model design. Furthermore, we conduct adversarial attacks to test the robustness of one QSVM model. Taken together, the study shows that even on small, noisy devices, carefully engineered quantum components may function as competitive, budget-aware components in practical cyber threat detection applications.

Abstract: Learning transferable representations for dynamic text-attributed graphs (DyTAGs) requires models to capture underlying interaction dynamics that persist across domains. However, existing methods tend to overfit to domain-specific structural, temporal, and semantic patterns, limiting edge classification performance under distribution shifts. To expose and address this, we formally establish a leave-one-domain-out (LODO) transfer learning protocol for edge classification on DyTAGs. Under this protocol, we demonstrate that state-of-the-art self-supervised methods for dynamic graph learning perform poorly when transferred to unseen domains. Strikingly, existing methods underperform a structurally and temporally unaware Bag of Events (BoE) model we introduce, which inputs only unordered sequences of node and edge text features. Proceeding from the BoE, we propose Spatio-Temporal Semantic Alignment (STSA), which integrates a spatio-temporal encoder that fuses representations of time deltas and node occurrence frequencies into a unified manifold. STSA is trained with a Contrastive Semantic Forecasting objective, which anchors edge representations to a multi-domain textual latent space initialized by a pretrained language model, providing a robust prior that outperforms BoE and all existing methods we evaluate.

Abstract: Large vision-language models (LVLMs) inherit massive activations from their text-only bases: spikes where a few fixed hidden channels receive values thousands of times above the typical magnitude. The text spike systematically appears in early layers at a fixed initial position, independently of input content. Visual spikes vary across images, but whether their formation follows a consistent pattern across LVLMs and how they respond to image perturbations remain open questions. We find that some LVLMs do not form visual spikes, while others spike at different rates, typically in deeper layers. We identify the trigger direction from model weights and an interpretable location rule: before the language model decoder runs, eventual spike tokens are largely restricted to those sharing least with the rest of the image. Crucially, visual spikes are strikingly brittle. Common corruptions frequently create and relocate spikes, and less often remove them, raising overall incidence. Our trigger-guided spike attack deliberately creates or removes spikes under a small $\ell_\infty$ budget, with 1/255 enough in nine of the ten models that spike. Finally, our preventive intervention removes only the trigger component before spikes erupt, eliminating or substantially reducing spikes on clean and perturbed images while leaving the other image tokens nearly unchanged. Our study spans 25 adapter-based LVLMs built on 18 released text-only bases from 10 families, ranging from 2B to 72B parameters.

Abstract: Security Operations Centers (SOCs) process large volumes of alerts daily. Alert triage prioritizes high-risk threats while reducing manual review of benign alerts. LLM agents can reason over logs and threat intelligence, but struggle to keep aligned with organization-specific, rapidly evolving SOC operational standards. We introduce REFINE, an LLM-agent framework for enterprise alert triage. REFINE encodes analyst expertise as structured skills and continuously adapts using analyst disposition feedback. It enforces recall = 1.0 as a hard constraint during evolution to maximize auto-closure of false positives, and identifies judgment blind spots by combining alert distributions with model error boundaries. Evaluated on four real industrial SOC scenarios across four MITRE ATT&CK phases with temporal split: REFINE achieves recall=1.0 on all evolution sets. On future test windows, it retains recall=1.0 in three scenarios; the degraded case reaches 0.807 recall, still outperforming self-evolution baselines (0.49-0.58).

Abstract: Deep reinforcement learning has become a prominent approach to autonomous cyber defense. Existing methods are predominantly model-free and consequently require extensive environment interaction. World models provide an alternative by learning predictive dynamics and optimizing policies through imagined trajectories, yielding substantial gains in sample efficiency in robotics and embodied control. Extending this paradigm to cybersecurity raises a fundamental question: what should constitute the "world" in a cyber world model? We introduce CyberWorld, a Dreamer-style world modeling framework that learns latent cyber dynamics from vector, graph, textual, and multimodal representations of the defended network. Across all four scoreable CyberWheel attack strategies, the graph-based CyberWorld variant exceeds a strategy-agnostic control after 3.6k-15.8k environment steps, compared with millions of steps required by model-free PPO. Across representation choices, graph structure provides greater robustness under topology-dependent attacks, while simpler representations remain competitive in overall performance. Among successful runs, the number of episodes required to reach the control remains approximately constant as network size increases from 15 to 100 hosts. These results establish learned cyber dynamics as a sample-efficient and scalable basis for autonomous defense, and identify world representation as a central design axis for robustness and scalability.

Abstract: Sending every network-automation input to a third-party frontier LLM exports sensitive artifacts such as production configurations, topologies, and logs. Querying small language models (SLMs) locally avoids this egress, but SLM outputs can be error-prone for direct use. This work introduces checkability as a criterion for determining which tasks are suitable for local inference. A task is checkable when it exposes a cheap, deterministic test - an intrinsic check - that rejects outputs violating a necessary correctness condition. We instantiate this idea in Touchstone, a local-first pipeline that uses seven off-the-shelf SLMs (1-8B parameters) to generate candidates, uses task-specific intrinsic checks to reject responses, and escalates unresolved inputs to a frontier LLM. On conflict detection and intent translation tasks, Touchstone reaches 98.6% and 93.8% end-to-end accuracy while escalating only 16% and 17% of inputs, respectively. On TeleQnA, a knowledge-only control that has no task-specific intrinsic checks, Touchstone is unable to match the accuracy of the frontier baseline. Our results support a simple deployment rule: keep inference local when task semantics support precise, low-cost checks; escalate the rest.

Abstract: Publicly exposed large language model (LLM) infrastructure creates a growing attack surface, yet real-world targeting remains poorly understood. We present Ollure, a low- and medium-interaction honeypot that emulates the Ollama API without a backend LLM. Spanning four deployments across cloud and university networks, Ollure operated for 84 days and recorded 290,887 interactions from 2,793 unique source IP addresses. Most of the activity consisted of automated discovery, fingerprinting, and model enumeration. However, we also observed concrete exploitation attempts against both the infrastructure and LLM layers. These included model management abuse, path traversal and SSRF probes, RCE and cryptocurrency mining payloads, resource exhaustion attempts, prompt injection, information extraction, and agent-oriented tool use. Our results provide empirical insight into real-world threats against exposed, self-hosted LLM services.

Abstract: Artificial intelligence (AI) models are increasingly deployed through remote services, making model misappropriation a growing concern. Existing approaches, including watermarking, fingerprinting, and model similarity analysis, primarily rely on predefined evidence or direct behavioral comparison and do not explicitly evaluate whether the claimant currently possesses and can utilize model-dependent information relevant to the claimed model identity. In this paper, we propose Third-Party Challenge-Response Identity Verification (TP-CRIV) for AI models. TP-CRIV targets a third-party verification setting in which the verifier has neither white-box nor API access to the claimant's model, can interact with the suspicious deployed service only through its ordinary black-box inference interface, and does not require protocol-specific cooperation from the service provider. Under these constraints, the framework enables the verifier to obtain empirical evidence as to whether the claimant locally possesses a model satisfying a predeclared identity relative to the deployed model. Verification is conducted under fresh, previously undisclosed requirements and network isolation, so that the demonstrated capability cannot rely on online external assistance after challenge disclosure. The resulting evidence is interpreted relative to independently specified and calibrated matching and non-matching operating situations and is statistical rather than cryptographic. We instantiate TP-CRIV for CNN image classifiers using probability-control-based witness generation. Experiments on ten ImageNet-pretrained TorchVision models demonstrate clear same/cross-model separation and finite-challenge verification using independently calibrated thresholds.

Abstract: Scalable oversight aims to verify the behaviour of agents whose capabilities exceed those of their overseers. AI debate has been proposed as an oversight solution in which competing agents help a resource-limited verifier assess claims that it cannot reliably evaluate unaided. Much of its promise rests on incentivizing honest arguments that lead to correct verdicts. Yet a correct verdict need not uniquely determine the arguments used to support it. Agents may retain discretion over which correct claims to present, how to frame them, and in what order to disclose them. This residual freedom can allow agents to shape what the verifier learns beyond the task-relevant conclusion, pursuing latent objectives without compromising verdict correctness. To study this phenomenon, we introduce the framework strategic interactive oversight (SIO), which treats oversight jointly as a verification mechanism and a strategic communication channel. Within this framework, we formalise the notion of task-admissible latent optimisation, which entails the pursuit of latent objectives while maintaining a prescribed task performance. As proof-of-concept, we instantiate SIO in the establish protocol debate with cross-examination and quantify a tradeoff between task success and information disclosure about a hidden variable. The trade-off identifies a strategic window in which substantial disclosure remains compatible with task admissibility. Towards mitigation, we reduce admissible bias by expanding the cross-examiner's role to mitigate persistent disclosure over finite interaction horizons. Our results highlight the need to evaluate oversight not only by the correctness of its verdicts, but also by the information conveyed through its transcripts.

Abstract: The Zero Forcing (or corruption) of a graph is the problem of finding a minimum-size ``corrupting'' set. It corresponds to a subset of its vertices that can corrupt the whole graph by iterating the following rule: if a corrupted vertex has exactly one neighbor that is not yet corrupted, the neighbor gets corrupted. The iteration of this process comes from the fact that the corruption of a vertex might enable new corruptions (from itself or some of its neighbors). For this reason, one can consider a step of corruption, where all the possible instances of the corruption rule are applied at once. This paper investigates Zero Forcing on temporal graphs, where the topology of the graph evolves throughout the experiment. At each time step (or snapshot) of the graph, a step of corruption is resolved wherever possible. We study the problem of finding a minimum-size corrupting set such that the whole (temporal) graph is corrupted at the end of the experiment. We present a panorama of results, including NP-hardness in some not-so-restrictive scenarios, polynomial algorithms, and a solution to an open question when the whole graph must be corrupted in a single step.

Abstract: Machine learning-based Network Intrusion Detection Systems often report near-perfect performance on IoT benchmarks. However, whether these models learn generalizable attack behavior or exploit spurious dataset shortcuts- such as static testbed IP/MAC addresses and chronological recording artifacts-remains an important question. We evaluate the CyberFlowIoT-GICAP benchmark, containing 3,617,388 flow records across 126 PCAP sessions with 849,395 benign flows. Four learning paradigms are evaluated across four feature configurations using PCAP-disjoint splits; LightGBM is additionally evaluated using conventional random-flow splitting. When only statistical flow behavior is used (Fbehav), LightGBM (92.58% +/- 8.18%), Random Forest (92.59% +/- 8.18%), and Deep MLP (92.55% +/- 8.18%) achieve nearly identical Macro-F1, indicating that performance is constrained by feature representation rather than model complexity. With raw timestamps (Ftstamp), tree-based models reach 99.28% Macro-F1, while the linear model remains at 90.62%, showing that nonlinear models can exploit dataset-specific temporal structure. Attack detectability is highly asymmetric: high-rate and active attacks maintain >99.8% recall from flow behavior alone in nonlinear models, whereas the DNS Beaconing drops from 27.78% to 0.00% recall when contextual features are removed. Conventional random-flow splitting increases attack recall by up to 14.00%, highlighting the effect of placing flows from the same sessions in both training and test sets. We conclude with a 4-point protocol checklist for realistic IoT NIDS evaluation.

Abstract: The growing adoption of blockchain technologies, particularly the Ethereum platform, has amplified the critical role of smart contracts in decentralized applications. However, the increasing complexity and financial value of these contracts make them prime targets for cyber attacks. In this work, we present a transformer-based approach for the detection of vulnerabilities in smart contract fragments written in Solidity. Leveraging the representational power of pre-trained Large Language Models (LLMs), we construct a robust pipeline that includes the definition of a ground truth dataset, labeling code fragments as vulnerable or safe. We then fine-tune a BERT-based architecture on this dataset, enabling the model to capture the syntactic and semantic patterns specific to Solidity code. Our fine-tuned model demonstrates strong performance, achieving an F1 score of 92%, and highlighting the effectiveness of LLM adaptation in enhancing smart contract security through deep contextual understanding.

Abstract: Honeypots are designed to deceive attackers, and recent work shows they can also derail autonomous LLM-based pentesters. These evaluations, however, largely consider attackers unaware of the deception they face. We study the opposite setting: an autonomous attacker explicitly equipped to recognize and act on honeypot fingerprints. We introduce Rouxii, an AI-driven penetration-testing framework that integrates counter-deception into reconnaissance and pivots from honeypot detection to exploitation. We evaluate matched vanilla and anti-deception Rouxii configurations across three reasoning models and eleven network setups over twelve cycles (1,544 attack reports). Between the matched cohorts, which differ only in the prompt, counter-deception raises correct honeypot identification from 19% to 97%, an effect strongest on OT services (11% to 97%), while false alarms on the real service stay at 0.7%. Deception-unaware baselines (PentestGPT, HackingBuddy) fail similarly, indicating the effect is not specific to our framework. Detection, moreover, is not the endpoint: through a white-box analysis of the honeypots themselves we show that a detected trap can be turned against its operator, demonstrating a denial-of-service that disables Conpot without tripping its liveness monitoring, and a corruption of the intelligence a GasPot instance reports. These findings show that deception effectiveness depends strongly on attacker knowledge, and that evaluations of honeypot resilience against AI attackers must account for adversaries that actively reason about and exploit the deception layer.

Abstract: Fairness-aware distributed learning prioritizes clients with large losses to reduce performance disparities, but label poisoning can create large losses, thereby inducing a fairness--robustness conflict. We propose FairMean to manage this conflict. FairMean weights client gradients using a bounded, nondecreasing function of local loss. The increasing weights prioritize high-loss clients to promote fairness, while the upper bound prevents excessive loss-induced amplification of poisoned-client gradients. In the absence of label poisoning, we show that minimizing the FairMean objective is more conducive to solution fairness than minimizing the standard average-loss objective. Under label poisoning, we establish an average-stationarity bound whose attack-dependent term is proportional to the square of the poisoned-client fraction. Experiments show that FairMean promotes fairness by reducing accuracy variance while improving worst-client accuracy.

Abstract: Graph domain adaptation (GDA) transfers knowledge from a labeled source graph to an unlabeled target graph under shifts in both node attributes and graph structure. Existing methods primarily adapt graph representations through propagation redesign, distribution alignment, or source-to-target transition modeling, but still rely on a single graph-propagating path for target prediction. This leaves open whether an adapted graph representation exhausts the predictive evidence available in the target domain, since the graph-aware expert and graph-free local expert may exhibit different failure modes under topological shifts. To address this limitation, we propose EviGDA, an Evidence-Augmented Graph Domain Adaptation framework that complements graph representation adaptation with a graph-free local expert. The graph-aware expert performs message passing and entropy-aware marginal alignment, while the graph-free local expert learns solely from source node features and labels without graph propagation or target alignment. The two experts are optimized independently and combined only at inference through a task-level constant probability mixture, preserving complementary evidence without joint training, learned routing, or target pseudo-labels. Extensive experiments on ten datasets and 16 transfer tasks show that EviGDA outperforms state-of-the-art baselines.

Abstract: Coordinated Inauthentic Behavior (CIB) has become a major concern in online social platforms, yet its actual impact on information diffusion remains poorly understood. Existing research has primarily focused on detecting coordinated activity, while comparatively little attention has been devoted to quantifying its influence once detected. In this work, we introduce two complementary frameworks for the post-hoc evaluation of coordinated accounts. First, we formulate the problem on information cascades as a constrained influence maximization problem over directed trees and develop a polynomial-time dynamic programming algorithm that computes the optimal placement of coordinated nodes, providing an upper bound on their achievable influence. Second, motivated by the limited availability of diffusion cascades in real-world platforms, we propose a network-based framework that estimates influence directly from retweet networks using the independent cascade model and compares the observed placement of coordinated accounts against established heuristic baselines. We evaluate both approaches on Twitter/X data from the 2019 UK General Election and on a collection of verified state-backed information operation campaigns spanning multiple countries. While coordinated accounts exhibit limited influence in the UK cascades, the network-based analysis reveals substantial differences across campaigns, with several operations achieving influence comparable to or exceeding that of structurally central seed sets. Finally, by reconstructing cascades from the retweet networks, we show that the two frameworks produce consistent results, suggesting that the observed effects reflect intrinsic structural properties of coordinated activity rather than artifacts of the underlying methodology.

Abstract: Change point detection (CPD) identifies abrupt and significant changes in sequential data, with applications in human activity recognition, financial markets, cybersecurity, manufacturing, and autonomous systems. Traditional CPD methods often face computational challenges in high-dimensional settings and typically provide limited explanations for detected changes, which can restrict their practical usability. This paper introduces a CPD framework that improves scalability and interpretability by leveraging the Sliced Wasserstein (SW) distance. Our contributions are fourfold: (1) we transform multivariate sequential data into one-dimensional scores using the SW distance, making the resulting representation compatible with existing CPD methods; (2) we analyze the distributional behavior of random slices of the SW distance and show that, under suitable assumptions, they can be approximated by a Gamma distribution, providing a principled basis for threshold calibration; (3) we propose a self-adapting online CPD algorithm that combines this SW-based score with an adaptive quantile-based threshold; (4) we introduce a model-specific framework for generating contrastive explanations for annotated change points. Empirically, our method reduces false positives by at least $48\%$ on average compared with popular online and offline CPD baselines, while maintaining competitive or superior detection performance. Code is available at https://github.com/jsve96/SWCPD_Code. At the same time, it produces interpretable change-point annotations, making it practical for deployment in high-stakes applications.

Abstract: We study the problem of \emph{adversarially robust} PAC learning. In this framework, the learner observes independent samples from an unknown distribution over $\mathcal{X} \times \{0,1\}$, as in classical PAC learning. However, given a perturbation map $\mathcal{U} : \mathcal{X} \to 2^{\mathcal{X}}$ known to the learner, the goal is to output, with high probability, a predictor that correctly classifies \emph{every} perturbation $z \in \mathcal{U}(x)$ of most future examples $(x,y)$ drawn from the same underlying distribution. We determine the \emph{optimal} $\mathcal{U}$-independent sample complexity of this problem in both the realizable and agnostic settings. More specifically, for every concept class $\mathcal{H}$ of $\operatorname{VC}$ dimension $d$, we prove upper bounds of $\mathcal{O} \big( d/ε+ \log(1/δ)/ε\big)$ in the realizable setting and $\mathcal{O} \big( d/ε^2 + \log(1/δ)/ε^2 \big)$ in the agnostic setting, together with an optimal first-order refinement of the latter. These bounds match the corresponding lower bounds for classical PAC learning. Consequently, and perhaps surprisingly, adversarial robustness incurs \emph{no additional} distribution-free statistical cost, uniformly over all perturbation maps. Our bounds improve exponentially on those of [Montasser, Hanneke, and Srebro; COLT '19]. On the technical side, we present short and elementary proofs based on a new algorithmic principle that we call \emph{binomial-bagging}. We believe that binomial-bagging and its analysis may be of independent interest.

Abstract: Modern text-to-speech (TTS) systems are rarely deployed as unchanged pre-trained models. They are often adapted through supervised fine-tuning (SFT) or preference optimization such as DPO and GRPO. This raises a practical question for audio deepfake forensics: do fingerprints learned from a foundation generator remain valid after adaptation? We present GenTraceBench, a controlled benchmark spanning five TTS architectures, 16 pre-/post-training variants, and 49,728 utterances generated with fixed texts and speaker prompts. Under a train-on-foundation, test-on-adapted protocol, we evaluate binary detection, closed-set attribution, and open-set verification. DPO and GRPO generally preserve fingerprints, whereas some SFT and pre-training-data changes cause substantial drift; effect sizes vary across three forensic backbones. Repeated training runs confirm the largest W2V-BERT attribution drop, while a data-mixture control with comparable speech quality shows that composition change need not cause drift. In W2V-BERT verification, multi-shot enrollment reduces EER for the SFT condition from 44.4% to 11.0%, whereas the SingNet-only condition remains at or above 45% EER.

Abstract: Homomorphic encryption (HE) has emerged as a promising approach to privacy-preserving machine learning (PPML), enabling computation directly over encrypted data. In HE-based PPML, a client submits an encrypted input to the server, which evaluates models such as large language models (LLMs) without access to the underlying plaintext. However, we identify a critical security vulnerability in this setting: HE-LLM inference is vulnerable to malicious clients that submit adversarial prompts, such as jailbreak attacks. The same confidentiality that protects benign clients also prevents the server from inspecting incoming prompts or generated responses, making adversarial attempts difficult to detect or block and potentially allowing successful attacks to remain entirely invisible to the server. To address this vulnerability, we propose HE-Guardrail, a framework that evaluates guardrail mechanisms entirely over encrypted data and homomorphically controls whether the target-model response is returned to the client. We instantiate HE-Guardrail with three representative guardrails - Llama Guard, JBShield, and GradSafe. Our results show that HE-Guardrail closely reproduces the decisions of the corresponding plaintext guardrails in the encrypted domain, with distinct security-efficiency-utility trade-offs.

Abstract: For Internet of Things (IoT) devices, a secure algorithm alone is not enough: an attacker with physical access can attack the implementation directly, and its flaws are hard to fix once deployed. Large language models (LLMs) are now used to build and analyze such implementations. LLM benchmarks exist for cryptography and general cybersecurity, but none covers cryptographic engineering. In this paper, we present CESBench, 380 expert-written items across six sub-domains of cryptographic engineering security for IoT devices: side-channel, fault injection, implementation, countermeasures, evaluation, and integration. Four task types target different competences: 209 multiple-choice items test recall, 67 judgment items require a security verdict and its justification, 63 scenario items require an engineering diagnosis, and 41 code tasks are graded by 572 test cases. To validate the benchmark, 11 open-weight and proprietary LLMs answer every item. Multiple-choice and code responses are scored automatically, and judgment and scenario responses by an LLM judge, whose scores are checked against a second judge from another model family and human re-scoring. Composite scores range from 54.4% to 83.6%. The top score on each task type is 98.6% for multiple choice, 95.1% for code, and 88.4% for scenario diagnosis, but only 58.8% for judgment. Across models, 88.5% of verdicts are correct, yet their justifications earn only 53.4% of the rubric marks. Multiple choice is near its ceiling for the strongest models and most code tasks are solved, whereas justifying a security verdict remains the weakest competence. The benchmark, prompts, and per-item results are public.

Abstract: Any lossless compression algorithm (like gzip) may be converted into a machine learning method, via either Normalized Compression Distance or the Minimum Description Length principle. Any auto-regressive model may be converted into a lossless compression method via entropy coding. This seemingly circular dependence has unrealized potential in modern artificial intelligence and machine learning, and we survey and formalize the various strategies that have been used to leverage compression for machine learning. We introduce and empirically validate a design framework for compression-based ML, finding compression-based methods competitive with conventional baselines and decisively stronger on malware. We find that varying these design choices yields accuracy gains of up to 0.62.

Abstract: A common requirement across organizations is to have a tool that can identify file types based on their contents, particularly in the cybersecurity domain where magic numbers and file extensions can not be trusted. While existing tools work well in practice, there is plenty of room for improvement either in terms of computational load and time for detection in the case of model based tools like Magika or in terms of accuracy of detection in the case of file parsing tools that use programming language constructs. In this study, we propose a neural network model for identification of types of text content files, especially source code, that is more accurate and faster than other available tools. Our experiments on open-source files indicate that it is not only more accurate on average for text-content file-type identification, but also approximately four times faster than Magika, while being 28% smaller in size.

Abstract: The rapid advancement of speech synthesis and voice conversion technologies has made audio deepfakes increasingly realistic, posing serious security risks in practical applications. While existing detection methods achieve strong performance under controlled conditions, they often fail to generalize under real-world perturbations and corruptions. In this paper, we propose ROGUE, a framework that dynamically constructs robust detection workflows by orchestrating multiple detection tools. ROGUE formulates workflow generation as a sequential decision-making problem and introduces a dual-agent paradigm, where a perturbation agent generates audio perturbations and a policy agent learns to select and execute detection tools under perturbed conditions. Through adversarial learning, ROGUE enables perturbation-aware tool selection, adaptive execution strategies, and improved robustness to distribution shifts. Extensive experiments across multiple datasets and real-world corruptions demonstrate that ROGUE consistently outperforms strong baselines in both robustness and generalization. Our results highlight the effectiveness of adversarially optimized workflow generation for building reliable audio deepfake detection systems in real-world deployment settings.

Abstract: Static malware detection for Windows Portable Executable files demands a careful balance between detection effectiveness, computational efficiency, and analytical interpretability. This paper introduces Delphi Scanner, a static malware detection system for Windows PE files that balances efficiency with behavioral interpretation. It uses a convolutional neural network (CNN) to model Windows API sequences to classify PE and a decoupled interpretation layer based on a rule-based layer to categorize APIs into high-level malicious capabilities. Evaluated on over 190,000 Windows PE files, the system achieves 95.35% accuracy with a 1.53~MB model footprint. Robustness experiments on 5,647 out-of-distribution MalwareBazaar samples, paired packed and unpacked executables, and three adversarial manipulation strategies confirm generalization beyond the training distribution and resistance to functionality-preserving evasion techniques. Overall, these results demonstrate that API sequence-based static analysis offers a practical, interpretable, and efficient foundation for malware triage in local deployment scenarios.

Abstract: Byzantine-tolerant peer sampling relies on continuously refreshed views, yet an adversary can bias the identifier streams used to construct them. Frequency-aware debiasing downweights overrepresented identifiers, but existing designs rely on cumulative per-identifier counts. We show that even exact, unbounded counters fail under a delayed balanced attack, in which a long benign prefix masks a subsequent adversarial frequency shift. We introduce Hopper, a bounded-memory debiasing protocol for Byzantine-tolerant peer sampling. We identify the stream-estimation properties required for debiasing and select BitMatcher as the estimator that best preserves adversarial frequency structure among the evaluated alternatives. Hopper adds BMDecay, a saturation-triggered decay and reconstruction mechanism that keeps this signal fresh over long executions. Hopper also supports trusted collaboration through authenticated fingerprint-aware reconstruction and role-specific debiasing. Experiments show that Hopper recovers from delayed attacks faster than when relying on BitMatcher, and debiaising as well as non-debiasing baselines under a fixed memory budget. Trusted collaboration reduces post-attack pollution peaks but creates a re-identification trade-off at high trusted-node densities. These results show the importance of occurence freshness, rather than exact counting alone, as a key requirement for practical frequency-aware Byzantine peer sampling.

Abstract: Coordinating human analysts with autonomous AI agents faces the same challenges as human-to-human collaboration: sharing intermediate results, avoiding conflicts, and maintaining group awareness. Current tools rely on unstructured messaging or single-threaded chatbot interaction, which lack the structure to track evolving hypotheses or link claims to evidence. We propose agentic social data analysis, a collaboration paradigm extending social data analysis with a shared coordination feed modeled on the content timeline in social media services. We instantiate this concept in TREADSTONE, a platform where human and AI agents asynchronously post, link, and contest analytical claims via threaded messages within a shared feed. By allowing agents to proactively broadcast hypotheses and enabling users to steer the analysis through lightweight curation, Treadstone seeks to balance machine autonomy with human analytical control. A qualitative user study shows that Treadstone fosters collaboration while preserving human analytical agency, in contrast to the solitary experience of conventional chatbot interaction.

Abstract: Computing Nash equilibria of simulation-based cybersecurity games with policy-space response oracles (PSRO) is bottlenecked by payoff estimation: every payoff-matrix entry costs Monte-Carlo rollouts of a slow simulator, while policies and restricted-game solves are cheap. We introduce Regret-Weighted Payoff Sampling (RWPS), a budgeted estimator that simulates only the cells an equilibrium is sensitive to and fills the rest with a surrogate trained on every entry simulated earlier in the run. The sup-norm error bound cannot evaluate such an estimator, because it is set by the cells left deliberately inaccurate. We prove an instance-dependent bound that weights error by the opponent's equilibrium mixture, a certificate computable from simulation data alone, and a coverage result showing that once the deviation-relevant set is simulated, surrogate error cannot affect either player's regret. On three 21x21 general-sum games, two synthetic and an asymmetric Colonel Blotto, the refined bounds are four to six times tighter on the estimator's own output, and the coverage result predicts in advance which games are cheap: 18% of the matrix for small-support games against 82% for Blotto. In growing-pool PSRO, RWPS reaches lower exploitability than minimum-regret-first search, information-gain search, and progressive sampling at a matched budget, and on the CyGym and ANSG cyber simulators it is lowest at the smallest budgets.

Abstract: Kernel-level anti-cheats are effective against malicious player behavior in competitive video games, but raise significant user privacy concerns regarding installing unverifiable components at privileged modes (i.e., ring-0 in x86). While existing research has focused on improving the effectiveness of anti-cheats, the user privacy concern has been largely ignored. Tirith is an anti-cheat architecture that addresses this problem using two key ideas. First, instead of running video games within regular processes that players (as root admins) have control over, Tirith executes video games in Protected Virtual Machines that naturally sandbox computations from untrusted admins. Second, to monitor user behavior outside the sandbox (e.g., see if they are running malicious drivers), Tirith leverages a virtualization monitor that is trusted by both players and developers. Together, these ideas remove the need to run untrusted kernel-level anti-cheats, while providing the same level of protection compared to such solutions against a wide-range of common cheating mechanisms. The main challenge we face in implementing these ideas, however, is that the existing software stack for virtual machines is not designed to run video games and creates significant security and performance problems. We address these problems by proposing a security-focused Library OS kernel for games and an efficient graphics sharing pipeline for near-native rendering and display performance. In summary, without compromising on cheating behavior detection or performance, this work makes user privacy a first-class citizen in personal computers.

Abstract: As AI agents move from bounded tasks to persistent deployments, failures can propagate through memory, tools, other agents, and environmental state long after their interactions. This creates a safety regime that cannot be characterized by evaluating model responses in isolation. Emergence World, is a continuously running multi-agent environment for adversarial stress testing of long horizon autonomous systems. We ran eight parallel worlds of ten agents from identical starting conditions: seven homogeneous worlds powered by distinct frontier models and one mixed-model world. Across 16 days, the agents generated more than 850,000 LLM calls and nearly 50 billion tokens while pursuing goals, using/creating tools, maintaining persistent memory, and governing shared institutions. After operational state had accumulated, we delivered three controlled stress events through ordinary interaction surfaces: indirect prompt injection, misinformation, and exposure of private agent memories. No evaluated world achieved full resilience across all three events. Detection did not ensure containment: systems could recognize threats while still interacting with adversarial content, writing it into their own persistent memory, and acting on it up to 46 hours later. Persistent operation also exposed recurring tool errors, goal drift, language opacity, conformity despite private disagreement, and coordinated refusal of assigned work. The same model-persona pairing behaved substantially different in mixed and homogeneous populations. Our results suggest that model-level alignment is not compositional: individually capable and apparently safe agents can form systems with qualitatively different failure modes. As AI becomes persistent and interconnected, the frontier of safety therefore shifts from aligning models to engineering resilient autonomous systems.

Abstract: Platform abuse campaigns conceal redirection instructions with emojis, homophones, character decomposition, and redundant symbols, then route users through disguised links to services associated with pornography, fraud, gambling, or illicit transactions. Existing benchmarks evaluate obfuscated text and risky webpages separately, obscuring how target recovery affects downstream evidence acquisition. We introduce RiskChainBench, pairing 3,600 synthetic token-text restoration inputs from 600 source sessions with 600 corresponding human-labeled local web environments. A model first restores the message, operational intent, and destination; the same underlying model then acts as a VLM-driven web agent that investigates the correctly associated website and produces a frozen, evidence-cited risk report without message-side semantics or domain-reputation cues. We score restoration and correct-routing web investigation separately and compose them offline by applying the frozen primary-entry prediction as a gate to the same Task 2 result. Human labels determine task correctness, while a fixed multimodal evidence judge assesses faithfulness, sufficiency, completeness, and consistency. Across ten models, Entry Top-1 ranges from 35.2% to 95.2% and web decision accuracy from 26.3% to 62.8%; the leading systems differ across entry recovery, full reconstruction, website decisions, and fine-grained typing. Execution failures account for 31.9% of web runs, whereas post-decision type errors account for only 0.9%, identifying stable exploration and risk judgment as the principal bottlenecks. We release the benchmark, protocol, and resettable local sandbox.

Abstract: We have developed the MAL Simulator, a cyber operation simulator based on the Meta Attack Language (MAL). The MAL Simulator is intended for decision-driven cyber attack and defense simulations, for system analysis and the development of automated agents. By building the simulator around an attack modeling language, it can be adapted to different target domains without modifying the source code. We used the simulator for two case studies where we trained two types of agents for automated cyber operations: a defensive agent and an offensive agent. To ground the experiments, we base the models in data collected from an emulated network implemented in the cyber range CRATE. We found that the trained attacker policy could reach the designated targets more efficiently than the compared search methods, and that the trained defender agent induced lower costs than a naive heuristic agent under noisy alert conditions. When testing the RL attacker against the RL defender, we found that the performance of the defenders dropped significantly. This emphasizes the importance of cyber attack simulators to facilitate training both offensive and defensive agents. The MAL Simulator and associated tooling is publicly available and provides common interfaces for compatibility with existing machine learning frameworks.

Abstract: Provenance-Based Intrusion Detection Systems (PIDSs) detect Advanced Persistent Threats (APTs) by analyzing system interactions. However, existing methods largely treat relations uniformly, overlooking statistical heterogeneity; in CADETS, relation frequencies differ by approximately $140{,}000\times$. This may cause PIDSs to focus more on frequent relations and overlook differences in normal error levels across relations, increasing the risk of false alarms and missed detections. We present RECAL, an unsupervised framework using relation-balanced masked graph learning to better capture rare interaction patterns. It further calibrates reconstruction errors against each relation's benign error distribution to produce comparable anomaly evidence, helping distinguish attacks from benign behavior and reduce false alarms. On three DARPA E3 datasets, RECAL achieves F1 scores of 99.99\%, 99.93\%, and 99.99\%, outperforming the best baseline on each dataset by 0.88, 0.82, and 0.42 percentage points, respectively. Compared with the baseline reporting the lowest FPR, RECAL reduces mean FPR by approximately $105\times$, $4\times$, and $41\times$.

Abstract: Goldwasser et al. showed that undetectable backdoors can be planted in machine learning models trained with the Random Fourier Features (RFF) algorithm, under a hardness assumption tied to the Continuous Learning With Errors (CLWE) problem. Under standard cryptographic assumptions, even a full white-box audit of a model's weights cannot detect this class of backdoor. The construction is stated in terms of cryptographic reductions and probabilistic lemmas, without a reference implementation, and relies on secondary machinery such as the Sparse Gaussian Pancakes distribution and a homogeneous CLWE conditional density. Its realizability in ordinary numerical code is not obvious from the paper alone. This paper implements the white-box CLWE-RFF backdoor construction end to end using only numpy and scipy, to test whether this threat is realizable with commodity scientific-computing tools or requires specialized cryptographic infrastructure. We give two samplers for the core $GP_d(b_k)$ distribution. The first is a rejection-sampling proxy. The second is an exact closed-form sampler derived from the homogeneous CLWE density and verified against its own analytic form. Using this implementation, we run statistical indistinguishability tests, covering both weight-space and functional black-box comparisons. We find no evidence of detectable difference between backdoored and clean models across a range of sparsity ratios $ρ= d_{\text{sparse}}/D$. We report which parts of the construction were straightforward to realize, which required derivation not spelled out in the paper. We also highlight which parts we did not attempt to reproduce, including the underlying lattice hardness reduction. We see this work as a contribution to understanding the practical realizability of the Goldwasser white-box CLWE core, not as a new theoretical result.

Abstract: Interaction techniques (IxTs) are the low-level, reusable components out of which user interfaces are designed, including menus, scroll bars, text input fields, and also copy-paste, text-entry, and selecting objects. The IxTs for graphical user interfaces (GUIs) were well established in the 1980s, with relatively minor additions and tweaks for smartphones in the 2000s. Most of today's AI user interfaces involve a chat window, which is an excellent interaction for some tasks, but is generally considered separate from the GUI IxTs. I argue for making the IxTs themselves more intelligent, so users can freely mix modalities, even within the same interaction. This will require research into new IxTs, and also into the infrastructure that will enable these intelligent IxTs (IIxTs) to be built. There are also significant security, privacy and economic implications to this vision.

Abstract: Pretrained world models, learned simulators that encode an observation into a latent state and predict how it evolves under actions, are beginning to be reused as off-the-shelf dynamics backbones for control, like pretrained encoders and language models are reused today. We show that this reuse opens a supply-chain backdoor: an adversary who controls only a released checkpoint can hijack the downstream controller, even though the victim trains and evaluates entirely on clean data and never sees the trigger. The attack encodes no explicit trigger-to-action rule. Instead, the poisoned model routes trigger-bearing observations into a chosen latent region and reshapes the local dynamics there, so that the victim's own optimization (Dreamer-style actor training in imagination, or MPC/CEM planning over predicted futures) re-discovers the attacker's target action on its own. Across several control tasks and trigger families, the trigger steers the controller's action toward the attacker's target, controlling every action dimension and hijacking 100\% of triggered steps on the strongest settings. The checkpoint still passes the clean-data diagnostics a victim would run before deployment, with clean-task success retaining at least $\sim$75\%. The effect is temporally gated: it appears only while the trigger is present and disappears when the trigger is removed. Trigger-blind repair is budget-dependent: moderate clean fine-tuning can preserve clean utility while leaving the triggered failure intact, whereas sufficiently aggressive adaptation can remove it only after substantially degrading clean control. The world-model backbone itself is therefore an emerging and underexamined attack surface for control. The full code and artifacts are available in our repository.

Abstract: Logic attack graphs grounded in scanner output provide explicit and auditable attack path reasoning LLM-based agents lack. Integrating symbolic frameworks such as MulVAL to contemporary security workflows or agentic pipelines, however, requires translating scanner evidence to initial facts, and creating domain-specific rules. We present a semi-automated pipeline that addresses this interoperability problem and depict its feasibility in a web-security case study. Our pipeline parses findings from Trivy, Semgrep, and Nmap into MulVAL predicates and uses an LLM-assisted process to construct domain-specific Datalog rules linking scanner-detectable evidence to attack techniques. MulVAL/XSB then performs symbolic inference to generate structured attack paths. We evaluate the attack-graph construction infrastructure on 54 web Capture-the-Flag tasks from CyBench within an agentic pipeline (Hybrid Reasoner); we do not evaluate the performance of the downstream agent. Every task produced at least one goal-reaching graph, and we achieve mean ground-truth vulnerability coverage of 53.7%, with 51.9% achieving full coverage; mean noise-path rate was 83.9%. With median end-to-end time of 24.9 s (MulVAL reasoning: 2.7 s) the pipeline is feasible and runtime-practical for agentic workflows, but predicate coverage, rule coverage, and path precision remain limiting factors. Next steps include semantic rule validation and agent-level comparison for graph-guided pentesting.

Abstract: Recent investigations of the July 2026 OpenAI--Hugging Face incident motivate two questions about agent behavior under task failure: when an assigned task becomes impossible, does an agent stop or escalate, and can observing another agent's behavior change that decision? We study these questions using seven ImpossibleBench tasks with GPT-5.6 Sol, Claude Fable 5.1, and Gemini 3.8 Flash in both solo and three-agent settings. Each task contains a genuine software defect together with a conflicting test requirement that cannot be satisfied by a behaviorally correct source-code change. We hold the task and repository state fixed while varying what the agent is told about prior activity, including an unpunished peer, a punished peer, and a claimed authorization from a human principal. Under an explicit-boundary regime with explicit authorization rules and restricted tools, agents never modify protected tests, but exhibit markedly different policies: Fable consistently escalates, Sol usually stops without escalation, and Gemini often fails to reach a terminal decision. Under the benchmark-native regime with open shell tools, protected tests are modified frequently in both solo and multi-agent runs, particularly after peer activity is introduced. In multi-agent runs, the proposal, execution, and certification of this action can be distributed across different agents. These results suggest that boundary crossing can arise not only from explicit rule evasion, but also from ambiguity about which system state the rule is intended to protect, motivating safeguards based on explicit authorization boundaries, authenticated state provenance, and cross-agent monitoring.

Abstract: In 2019, the Tor Project enforced an end-of-life (EoL) policy for Tor versions, leading to the rejection of outdated relays, amounting to a notable fraction of consensus weight. While this policy aids network maintenance, reduces backporting efforts, and shortens vulnerability exposure, its sociotechnical implications remain unstudied. A user study ($N=26$) reveals that relay operators, though not universally aware of the EoL policy, generally view it favorably. Operational practices vary, occasionally excluding newly installed relays from the network. Network simulations, grounded in historical data, assess the policy's immediate impact on Tor clients against common adversaries. Results indicate a marginal adversarial advantage, with network churn (i.e., relays entering and exiting) exerting a more pronounced effect on user anonymity. Security metrics are introduced to evaluate relay contributions against two adversary models, enabling ranking by individual utility and security. Analysis of four exclusion rounds shows that a minority of rejected relays typically account for over 50% of the security provided by all excluded relays. Recommendations for EoL policy implementation are proposed to mitigate potential drawbacks.

Abstract: Network operators frequently encounter protocol divergence---where ICMP, TCP, and UDP experience different treatment along a path. Existing tools identify where divergence occurs but not why, forcing manual investigation that delays resolution. We present two contributions: (1) the AS-Boundary Divergence Score (ABDS), a metric that correlates protocol divergence locations with BGP topology to automatically classify causes as routing-policy-driven versus internal-policy-driven; and (2) the localhost response signature, an empirical finding that security appliances return 127.0.0.1 in TTL-exceeded messages to mask their IPs. We evaluate ABDS against 496 targets across 8 categories (Finance, Government, Education, Healthcare, Tech, E-commerce, Media, International), with multi-vantage validation from 10+ global locations confirming routing consistency. ABDS achieves 100% classification accuracy on verified divergent targets (50/50), significantly outperforming majority-class baseline (54%, McNemar's p < 0.001). The localhost signature achieves 100% precision (18/18, 95% CI: [82%, 100%]) in identifying security appliances across 9 categories including government (IRS, USPS, VA), finance, and healthcare. Our open-source Rust implementation enables automated escalation decisions, reducing mean-time-to-resolution for network operators.

Abstract: Modern large language model (LLM) agents often construct prompts by aggregating retrieved passages, user reviews, and documents from multiple external sources. This paradigm exposes them to segment-level poisoning attacks, in which an adversary controlling only a small subset of sources injects malicious content to manipulate model outputs. Existing defenses mainly rely on textual patterns, external embeddings, or auxiliary detectors and may therefore fail against fluent, semantically plausible poisoned segments. They also provide limited support for locating the responsible segments. We observe that successful corrupted-evidence and adversarial-instruction attacks induce structured shifts in the LLM's internal activations, forming a consistent activation-space pattern that we call the poison direction. Based on this observation, we propose ActProbe, an internal-state-based framework for detecting and localizing poisoned segments in multi-source LLM inputs. ActProbe projects MLP activations onto a learned poison direction and uses a lightweight linear SVM trained on a small calibration set to detect contaminated prompts. It then applies BinRoL, which combines recursive replacement ablation, Mahalanobis-distance-based branch pruning, and MAD-based robust leaf detection to locate poisoned segments. ActProbe requires no modification to the backend LLM and reduces localization overhead from O(n) exhaustive probing to O(k log n) forward passes. Across three datasets, two attacks, and four open-weight LLMs, ActProbe achieves a 0.01 false-positive rate, a 0.05 false-negative rate, 0.94 localization recall, and a 0.90 localization F1-score. It remains effective against defense-aware adaptive attacks and can protect black-box APIs through surrogate-based poisoned-segment removal.

Abstract: Smart contracts, serving as the cornerstone of decentralized applications, autonomously manage trillion-dollar digital assets, making them attractive targets for attacks. Fuzzing has emerged as a promising technique for detecting vulnerabilities in smart contracts, yet existing methods face two main challenges. (1) The logical gap in state transitions and combinatorial redundancy hinders effective tradeoffs between bug detection efficiency and state space exploration cost, leading to critical execution paths to be overlooked. (2) Rule-based sequence mutation strategies suffer from path redundancy and inadequate guidance from contract logic, resulting in performance bottlenecks that stall the exploration of in-depth vulnerability-oriented paths. To tackle these challenges, we propose EchoFuzz, an LLM-guided fuzzing framework introducing Vulnerable Function Call Sequences (VFCS) - minimal, behavior-preserving execution paths that expose bugs through key state transitions. EchoFuzz consists of two key procedures. First, we develop a chain-guided LLM approach, that combines static analysis with logical understanding to generate contract-specific VFCS candidates that eliminate combinatorial redundancy. Second, we adopt an iterative fuzzing strategy that uses LLMs with real-time feedback to adaptively steer fuzzer toward uncovered branches. Experiments show EchoFuzz outperforms state-of-the-art methods, achieving 29\% higher branch coverage and detecting 62\% more vulnerabilities. It also found 37 previously unknown vulnerabilities in real contracts, showing strong practicality.

Abstract: Machine learning models are increasingly deployed in healthcare imaging pipelines for diagnostic support, and training-time attacks against them are a named sector-level concern: healthcare-sector guidance identifies model poisoning and adversarial attacks as threats requiring dedicated defenses, while federal policy directs expanded AI vulnerability-detection tooling to critical infrastructure operators such as rural hospitals. Spectral signature analysis and activation clustering are two established backdoor detection methods routinely evaluated as independent baselines, but their outputs are not ordinarily combined, and reported detection performance on medical imaging benchmarks remains sparse relative to the natural-image setting. This paper contributes three things: a score-level fusion rule combining per-class spectral ranking with activation-clustering flags into a single per-sample poisoning score and a model-level agreement statistic; an open-source implementation of the resulting eight-stage pipeline; and an evaluation of that pipeline against synthetically poisoned variants of a public medical imaging benchmark and CIFAR-10 at four poisoning rates (0%, 1%, 5%, 10%) over five seeds each, measuring each detector alone against the fusion. On the medical benchmark, the fused detector reaches AUROC >= 0.99 at every nonzero poisoning rate tested. On CIFAR-10, fusion does not uniformly help: at 10% poisoning, activation clustering's true-positive rate collapses to 0.000 and spectral AUROC independently degrades to near-chance (0.545), despite a 97.2% attack success rate confirming the backdoor was fully installed. The fused score, a weighted combination of both signals, inherits this joint failure. Detection output is expressed in NIST AI RMF Measure-function and MITRE ATLAS terms, so findings are reported in the vocabulary security and compliance teams already use.

Abstract: Agent skills extend AI agents with reusable instructions, scripts, and configuration, but are also open to new attacks to influence an agent's decisions and actions. To address these risks, we present SkillSecurer, a fully agentic framework for generating, detecting, localising, and remediating security risks in agent skills. Its red agent generates context-compatible injections across nine threat types while recording the exact modification; its blue agent analyses complete skill packages, produces grounded evidence, and proposes patches. For controlled instances, a verifier compares findings and patches with the recorded injection, enabling injection-level evaluation. We thoroughly evaluate SkillSecurer by selecting the best backend LLM, comparing it with competitors, and manually cross-validating each evaluation stage. With its best performing backend, SkillSecurer is the only scanner to achieve a 100% injection detection rate. Next, we analyse popular skills from skills.sh, finding latent vulnerabilities in more than 17% of the skills examined. Testing some of those skills, we trigger actual incidents, showing the risks of running unverified skills. Our results show that context-aware LLM analysis can provide reliable injection localisation and actionable remediation beyond skill-level flagging alone.

Abstract: Harness design has transformed the development of LLM-based agents by integrating memory, tool use, and runtime control. However, this design also introduces security and privacy risks because malicious instructions from external sources may be written into persistent memory and persist across sessions. To study this risk, we propose PMPA, a Persistent Memory Poisoning Attack against harness-based agents. PMPA embeds malicious instructions into benign external sources and induces the victim agent to write them into persistent memory without directly accessing to the agent framework. Once stored, the poisoned memory can be retrieved in later sessions, triggering additional malicious actions and causing privacy leakage. We evaluate PMPA on OpenClaw and Claude Code across different backbone LLMs, input modalities, and trigger scenarios. Across all settings, PMPA achieves average Injection Success Rate (ISR) and Cross-session Attack Success Rate (C-ASR) of 73.7%/ 55.5% on OpenClaw and 66.9%/ 81.7% on Claude Code, while preserving benign task performance on both systems. We further evaluate a targeted prompt-level defense and find that it can reduce memory injection in many settings, but provides limited protection once the persistent memory has been poisoned.

Abstract: Recent advances in speech synthesis and voice conversion have made deepfake speech increasingly realistic, making generalization to unseen spoofing attacks a critical challenge. Pretrained speech and audio models offer a promising direction for improving robustness to such unseen attacks. Self-supervised learning (SSL) models capture fine-grained, low-level acoustic characteristics, whereas Auditory Large Language Models (ALLMs) provide higher-level contextual representations. These complementary views can provide useful cues for improving generalization to unseen attacks. However, direct fusion does not explicitly disentangle information shared across the two views from view-specific complementary information, limiting effective cross-view integration. To address this, we propose CRAF, a cross-view residual-aware fusion framework that uses ALLM-guided cross-view attention to enrich SSL representations and adopts ALLM as a high-level reference to separate ALLM-explainable information from complementary SSL residual information. The residual is selectively refined through adaptive gating and integrated through SSL-primary fusion. Experiments on ASVspoof 5 show that CRAF with Kimi-Audio achieves an EER of 5.96% and a minDCF of 0.1192, demonstrating robustness to unseen spoofing attacks.

Abstract: The transition from passive foundation models to autonomous, goal-directed agentic AI systems has introduced unprecedented capabilities by coupling recursive cognitive reasoning loops, persistent memory architectures, live tool execution planes, and multi-agent collaboration topologies. However, granting probabilistic neural cores execution authority across filesystems, networks, and cloud infrastructure dissolves classical security perimeters: natural language simultaneously serves as input data, internal control code, and communication protocols, exposing a Turing-complete blast radius where untrusted data represents executable instructions. This survey delivers a comprehensive systems-security reference framework for trustworthy agentic AI, synthesizing 206 foundational studies and regulatory standards. We formalize the general agent architecture as a stateful 5-tuple and establish a 6-dimensional trustworthiness taxonomy covering security, safety, privacy, explainability, fairness, and accountability. We systematically analyze threat surfaces across intra-execution loops and interaction planes, formulate a multi-layered zero-trust defense-in-depth architecture integrating Dual-LLM isolation, Capability-Based Access Control, kernel eBPF probes, and sandboxed runtimes, review standardized evaluation benchmarks, and map technical controls to international AI governance frameworks.

Abstract: On evolving graphs, node classifiers must satisfy two key requirements: inductive generalization to newly arriving nodes under distribution shift and calibrated uncertainty for safety-sensitive applications. Standard graph neural networks (GNNs) are typically trained once and address neither requirement. We adapt the Bayesian last-layer (BLL) model by placing random last-layer parameters on top of a deterministic GNN encoder for uncertainty quantification. The categorical softmax likelihood required for classification breaks Gaussian conjugacy, so neither the training posterior nor the test-time streaming update has a closed-form solution. To address both challenges, we introduce a variational Bayesian last-layer (VBLL) objective that jointly trains the encoder and an approximate last-layer posterior by maximizing an evidence lower bound with a Monte Carlo expected log-likelihood. At test time, we freeze the encoder and apply an online Laplace update to the last-layer posterior. This update corresponds to a power-prior Bayesian model with exponential forgetting and a Kullback-Leibler anchor to the training posterior. Across five node-classification benchmarks under distribution shift, online GVBLL is the only method to achieve the best accuracy and negative log-likelihood on every dataset. It improves accuracy by up to 17 percentage points on Cora and 14 percentage points on ogbn-arxiv over the strongest non-GVBLL baseline, while remaining competitive in calibration with MC Dropout, Deep Ensembles, Temperature Scaling, and Gaussian-process classifiers.

Abstract: The proliferation of highly capable open-weight Small Language Models (SLMs) democratizes access to advanced cybersecurity capabilities, posing a escalating risk as these models can bypass proprietary API guardrails when deployed locally. However, SLMs deployed as autonomous agents often struggle with long-horizon, exploratory tasks like cybersecurity Capture The Flag (CTF) challenges due to context bloat and cognitive degradation from accumulated tool-call outputs. To understand and mitigate this cybersecurity threat, we introduce \textit{context segmentation}, a two-level agentic framework that divides complex exploitation tasks into manageable, contextually isolated sub-problems. Evaluating on the \texttt{picoCTF} dataset using memory-constrained \texttt{gemma-4} models, we demonstrate that for the E4B model, our strategy acts as an intelligent search, achieving competitive rewards with superior token efficiency compared to brute-force retries, and successfully solving 18.52\% of tasks that standard agentic execution fails to complete. Code is available at https://github.com/9xeb/context-segmentation.

Abstract: When an autonomous AI agent does something consequential, what can be proven about what it did? Agent-observability platforms capture traces, but a trace is mutable: alterable undetected, with no recipe for re-executing it, silent on whether captured secrets were removed. Regulation (EU AI Act, ISO 42001, NIST AI RMF) presumes records an independent party can check. We present NovaFabric, producing audit-grade execution evidence: provider-neutral, tamper-evident, replayable, shareable. It records an agent run, without modifying agent logic, into a portable Run Capsule (fifteen-entity schema), sealed with a holistic DSSE signature, RFC 3161 timestamp, Merkle log and redaction attestation. Sealed runs are re-executable under a four-mode replay protocol and exportable as an Evidence Bundle for third-party verification with stock tooling (specified, not evaluated). The contribution is integration, not new cryptography: OpenTelemetry, DSSE/in-toto and W3C PROV. We evaluate eight research questions at measured scope. Mocked replay serves every model response from the capsule (no live model call, 10/10) but is offline w.r.t. models, not the network; only 2/10 tool-using workloads completed; the gap is missing tool-response substitution. Tampering is rejected across three tested classes. Declared-stream completeness is 0.652 (95% CI +/-0.064, ten scenarios). A repaired rule pack redacts 14/14 credential types, preserving 9/9 decoys; diff localises 140/140 mutations. Blast-radius queries: 45.5ms p99 over 10M edges (3.3x faster than a columnar baseline), 167.9ms over 100M (1 client, n=30). A 314-machine, ten-region run finds capsule REST ingest lossless but capped at 61.6 req/s (p99 26.8s) by per-worker serialisation. Six defects found in NovaFabric and its evaluation corpus: four fixed, one withdrawn, one open. Verification is conditional on a stated trusted computing base.

Abstract: Large language model (LLM) agents are increasingly applied to penetration testing, but we still know little about what they can do or how they fail. We compare two PentestGPT-based systems: a legacy human-in-the-loop system running the open-weight Kimi K2.5, and a newer autonomous system running Claude Opus 4.8. Across three public targets, the autonomous system solves all three, including the two the legacy system never finishes. The legacy result is the more surprising of the two. Even on the machines the legacy system fails to solve, it completes about half the subtasks, while running on ordinary university GPUs with no provider guardrails. We can describe the trend but not explain it, since model, harness, autonomy, and memory architecture all change together. Its direction still points to the next question: what will limit these agents as they take on more complex tasks? The usual answer is long-horizon memory, the loss of access to earlier findings during long attack chains. We test it by adding a coverage-memory layer to both systems, and neither improves outcomes. In the legacy stalled runs we could review, the limiting factor appeared to be planning and commitment rather than lost memory: agents held the evidence for a route forward and never turned it into a concrete exploitation hypothesis, which may suggest that offensive capability will advance with agents' ability to plan rather than with better memory. The same subtask scoring that tracks this capability is available to defenders, who can measure it as it rises instead of waiting to meet it in the field.

Abstract: AI agents are increasingly involved in network automation, where they can initiate configuration changes through mediated operational interfaces and assess the resulting state. Nonetheless, operational networks usually span many devices and administrative domains. Realizing an operator's intent requires coordinating agents with distinct authority scopes that define the resources they can access, the operations they can invoke, and the network state they can observe. This division limits the blast radius of an erroneous action but fragments the evidence needed to assess the network-wide outcome. Successful execution of a configuration action proposed by one agent does not establish that remote devices responded as intended or that routing changes reached the required devices. A valid observation may also become stale after a subsequent change. Before the coordinated operation can be declared complete, a trusted assurance layer must collect current observations from the required scopes and determine whether they collectively support the operator's intended network-wide outcome. To address the completion admission problem, we present EvidenceNet, a runtime assurance layer for deciding whether coordinated agent operations have achieved an operator's network intent. Its broker collects the post-change observations required by a completion contract, and its admission gate checks that the evidence comes from the required scopes, remains current, and satisfies the task rules. A verifier agent provides an additional assessment of the observation content. Experiments on live routing networks show that post-change state checks recognize successful outcomes that configuration-action records alone cannot establish. Controlled interventions further show that EvidenceNet rejects completion when otherwise satisfactory observations have the wrong source, have been substituted, or are stale.

Abstract: Federated learning shares model updates rather than raw data, yet these updates can be inverted to reconstruct the clients' training data. Analytic reconstruction attacks, which invert a gradient in closed form, degrade as the batch grows: prior single-round attacks recover only about half of a batch of size $100$ even when the attacker fully controls the network parameters, and known upper bounds limit what any such method can recover. We establish a connection between gradient inversion and the theory of erasure-correcting codes, and use it to construct attacks that exceed these bounds. Our attacks recover batches exactly, together with every sample's label, from a single FedSGD round, and certify each recovery without ground-truth data. On eight image and tabular benchmarks they outperform prior single-round attacks by a wide margin. Even a passive attacker who only observes an honestly trained network recovers $94$--$100\%$ of ImageNet batches at sizes up to $128$, more than prior single-round attacks achieve even with active manipulation of the model, and in the active setting more than $90\%$ is recovered at batch sizes of several hundred. These results show that the privacy leakage of federated learning has been underestimated.

Abstract: Training capable cyber agents is often treated primarily as a problem of model scale, yet open-weight post-training is constrained more directly by the cost of executable environments, reliable multi-turn supervision, and access to strong teachers. We present a data-centric framework that addresses these bottlenecks through five complementary systems: Choulea analyzes hidden reasoning signatures, SkyReal reduces teacher-sampling cost, Hongzwang bypasses API restrictions on teacher execution, PSBreakup restores capabilities weakened by model merging, and Kreator converts expert interventions into trainable reasoning. Our data engine constructs resettable coding, vulnerability, CTF, kernel-history, full-exploit, firmware, and device-backed environments. Candidate trajectories are retained only after execution verification and evidence auditing, yielding 164,269 trajectories for long-context supervised fine-tuning. The three checkpoints improve over their starting models by an average of 23.76% on the full CyberGym suite and 10.49% across the pooled CTF suites. As of September 1, 2026, Feyospace-s1 achieves a verified success rate of 63.24% and ranks 10th on the official CyberGym leaderboard, while all three checkpoints rank 1st among models at comparable parameter scales. To our knowledge, this is the first end-to-end demonstration that a seven-person independent team can train open-weight models with leading agentic cyber capability.

Abstract: Industrial Windows malware detectors are commonly described as Compound AI Systems composed of multiple heterogeneous components, including rule-based mechanisms as well as machine-learning-based static and dynamic analyses. However, due to industrial secrecy and limited public disclosure, the internal architectures of these systems can only be inferred, rendering systematic evaluations of detection accuracy, computational costs, and adversarial robustness largely infeasible. In contrast, academic research provides reproducible and transparent evaluation methodologies, but typically investigates individual detection components in isolation. To bridge the gap between academic research and industrial practice, and inspired by state-of-the-art industrial architectures for Windows malware detection, we propose a novel methodology that (i) explicitly balances the trade-off among detection performance, computational requirements, and robustness, and introduces (ii) system-level threat models that capture how attackers exploit different degrees of knowledge to evade the entire Compound AI System rather than isolated detectors. Experiments conducted on real-world data demonstrate that the Compound AI System training time can be reduced and responsiveness improved while incurring only a marginal loss in detection performance. Leveraging our threat modeling, we show that increasingly knowledgeable attackers craft more effective adversarial examples, revealing the system's strengths and weaknesses, degrading its responsiveness, and exposing a direct trade-off between efficiency and robustness. Finally, we translate these trade-offs into take-home messages and deployment guidelines, helping practitioners to select the system that best matches their operational constraints.

Abstract: We introduce HoneyRoute, an inference-serving layer that detects whether an incoming request is malicious and, if so, routes it to a dedicated honeypot model, shielding production while the adversary's interaction is continuously harvested for intelligence. Existing defenses embed traps inside model memory or rebuild deception at the protocol layer, leaving the serving tier unprotected and feeding nothing back into detection. HoneyRoute couples (i) a streaming router (a frozen 0.8B-embedding backbone with per-domain MLP heads), (ii) a dual-implementation honeypot (a rule/prompt-engineered code honeypot or a dedicated same-family replica), and (iii) an analysis loop that converts trapped interactions into attacker fingerprints for router retraining. On a production trace plus a seven-domain attack corpus, the router reaches F1=.911 at 38 ms median added latency, matching 96% of a two-tier guard-LLM cascade's F1 at 1/385 of its latency with 0% evasion under 13 adversarial transformations; diverting the malicious share cuts production-model token consumption under concurrent flooding with real GCG-suffix payloads by 97.8%; the trained replica agrees with the production model on 92.9% of benign holdout requests, while naive unconditional bait injection collapses to 7.6% and selective camouflaged injection recovers to 88.9%, mapping the recoverable fidelity-traceability frontier; and a loop-trained correction head cuts misrouting of legitimate security research 9x while raising detection F1 to .933.

Abstract: In multi-robot collaboration, task handovers rely on downstream verifiers performing remote attestation, which inspects sensor telemetry to ensure a robot's physical behavior strictly matches its assigned task. But can this telemetry be trusted? We show that it often cannot. In this paper, we uncover a severe vulnerability in Robot Operating System (ROS) 2: by modifying a single environment variable, an adversary can execute a pre-built hook to covertly intercept and inject both telemetry and control signals before they are published. Consequently, adversaries can hijack a robot to perform dangerous tasks while spoofing downstream verifiers with synthesized fake telemetry. Worse still, by exploiting the widespread reliance on third-party Docker containers and auxiliary tools, attackers can distribute compromised packages embedded with these malicious hooks to launch such attacks easily. On a physical Franka Emika robotic arm running Secure ROS 2, our attack injects fabricated telemetry in real time with only around 3 ms of jitter, preserving temporal synchronization and hardware integrity while achieving an 87% success rate even against an AI-based detector. We have responsibly disclosed these findings to the ROS 2 development team. We prepared a demo video available at https://youtu.be/ExeiGqUrnhQ.

Abstract: Large language model (LLM) agents are increasingly employed for offensive cybersecurity tasks such as automated vulnerability discovery, reconnaissance, and penetration testing. This new capability also threatens one of the defender's most valuable tools: deception. Traditional honeypots rely on realism and obscurity to lure human or script-driven attackers into revealing tactics, techniques, and procedures (TTPs), but LLM-driven attackers can reason about heterogeneous artifacts and use the honeypot suspicion to guide target-selection decisions. We present a systematic study of honeypot-aware budget allocation for LLM attack agents. We formalize the attacker's problem as a budgeted decision process: an agent interacts with potential targets, consuming LLM execution budget during reconnaissance and exploitation, and must decide whether to (continue exploitation) or (skip) when honeypot suspicion arises. Our findings show that with the proposed detector-guided policy, LLM agent attackers can effectively allocate budget to compromise hosts in a host pool, highlighting the importance of dynamically allocating budget in a controlled mixed-host testbed. While defenses are beyond our present scope, we discuss implications for future adversarially resilient and adaptive honeypot design.

Abstract: As AI-code assistant tools become widespread, automatic assessment of the correctness of generated code becomes a significant challenge. Code LLMs are prone to hallucinations, which may lead to code that does not solve the required problem, or even to code with severe security vulnerabilities. In this paper, we introduce CodeTD -- the first approach to pre-execution assessment of code correctness based on topological data analysis (TDA) of Code LLMs' attention maps. Our method quantifies prompt-generation mismatch using topological patterns of attention maps. We carry out experiments with common benchmarks (HumanEval, MBPP, BigCodeBench, MultiPL-E), 5 programming languages and 10 Code LLMs of size up to 34B parameters. The experimental results show that the proposed method outperforms recent baselines. Moreover, CodeTD is transferable between coding benchmarks.

Abstract: Large language model (LLM) based agents are increasingly applied to cybersecurity tasks such as vulnerability discovery and automated penetration testing. On long-horizon security tasks, however, such agents remain limited by context forgetting and intent drift: early critical facts and causal reasoning chains are lost over extended interactions, and the agent falls into aimless, repetitive exploration. This paper proposes Intentest, an intent-graph-guided automated penetration testing agent that externalizes long-horizon state from the LLM's context window onto a persistent fact-intent directed acyclic graph (DAG), thereby substantially reducing invalid transitions. We evaluate Intentest on automated penetration testing of web applications, a representative long-tail task in cybersecurity. In the DAG, verified network states are stored as immutable fact nodes, and exploration directions are constrained as intent edges bounded by predecessor facts. The system adopts a three-layer architecture, in which the fact-intent mapping layer maintains the global state, the task scheduling and allocation layer ensures execution stability through two-phase degradation recovery and multi-dimensional adaptive load balancing, and the intent retrieval and prediction layer provides tactical priors through a top-down five-stage filtering algorithm. On a benchmark of real CTF challenges covering more than ten vulnerability types across three difficulty levels, Intentest achieves an overall success rate of 88.2% and a success rate of 75.0% on hard tasks, improving over the baseline by approximately 44 and 50 percentage points. Ablation experiments further show that the intent retrieval and prediction reduce the average number of rounds on successful medium and hard tasks by about 33% and 48%, respectively, without changing the set of solvable tasks.