Papers

As Field-Programmable Gate Arrays (FPGAs) scale in multi-tenant cloud and edge-AI environments, the configuration bitstream has become a critical, yet opaque, security boundary. Existing hardware Trojan detection methods often rely on trusted design artifacts or computationally intensive reverse-engineering, introducing prohibitive latencies in dynamic, "just-in-time" reconfiguration workflows. This paper presents BLADEI (Bitstream-Level Abnormality Detection for Embedded Inference), a bitstream-level security framework designed for deployment-time screening of FPGA configurations without requiring source code, netlists, or vendor-specific tooling. BLADEI introduces a hybrid architecture that combines multi-scale byte-sequence learning with compact statistical representations to detect anomalous configurations directly from raw bitstreams. We implement the framework on a Xilinx PYNQ-Z1 system, demonstrating an end-to-end cloud-to-edge pipeline that enforces security prior to FPGA configuration. Evaluating across 1,383 bitstreams, BLADEI achieves a macro F1-score of 0.91. However, our systems-level characterization reveals a "preprocessing wall": software-based feature extraction accounts for 92% of the total 16.4-second latency, while model inference requires only 1.4 seconds. To address this bottleneck, we propose a streaming hardware-accelerated feature extraction engine designed for the FPGA programmable logic (PL). The evaluation shows that PL-based streaming engine can reduce feature-extraction latency to the millisecond range. This work positions bitstream-level screening as a first-class primitive and demonstrates that hardware-accelerated preprocessing is the key enabler for securing next-generation reconfigurable custom computing machines at line rate.

Autonomous AI agents are increasingly deployed on blockchain platforms, yet the design space that governs their interaction remains poorly understood. This convergence, where autonomous agents operate on and within decentralized systems, is a defining feature of the emerging Web~4.0 paradigm. This paper presents a Systematization of Knowledge organized around a bidirectional trust framework. In the B $\boldsymbol{\rightarrow}$ A direction, we examine how blockchain provides trust infrastructure for agents, spanning identity and account abstraction, permission and delegation, intent-centric execution, and tokenized agent economies. In the A $\boldsymbol{\rightarrow}$ B direction, we examine the reverse: how AI agents participate in core blockchain mechanisms including security auditing, consensus, and governance. A Trust Foundation of verifiable computation underpins both directions, with each primitive offering different trade-offs between trust minimality, computational overhead, and deployment readiness. We formalize the interaction as an Agent-Blockchain Interaction Model (ABIM), catalog 70 Ethereum EIPs/ERCs, examine 20 representative industry projects, and review 118 academic papers, applying a five-dimensional framework assessing Verifiability, Minimality of Trust, Expressiveness, Composability, and Maturity. Our analysis uncovers significant gaps: the agent-specific standards ecosystem is overwhelmingly immature, intent architectures lack formal analysis, and while isolated works have begun to explore AI participation in consensus and governance, a unified security framing that treats AI as a first-class actor at the protocol layer remains absent. We propose a three-dimensional taxonomy, identify nine concrete open problems, and highlight the sharpest research opportunities at this intersection.

The new wave of adversarial attacks that utilize gradient-related vulnerabilities in neural network-based classifiers makes Network Intrusion Detection Systems more open to such threats. Although state-of-the-art adversarial training methods have shown promising results in producing more robust classifiers, their interpretability and defense ability are limited due to their lack of understanding of how adversarial attacks propagate in different layers of network classifiers. In this paper, we present an insightful approach, called LARAR (Layer-wise Adversarial Robustness using Adaptive Regularization), that incorporates additional layer-wise vulnerability analysis and adaptive weighting in conventional adversarial training methods. Additionally, we utilize 'Auxiliary Classifiers' in our approach. LARAR provides interpretable layer-wise vulnerability scores, achieves a clean accuracy of 95.01%, and provides better robustness against adversarial attacks (FGSM, PGD, and transfer attacks) on the UNSW-NB15 dataset. Through the identification of vulnerable layers, the proposed framework reduces computational complexity and enables the early detection of adversarial samples, thus enhancing the effectiveness and interpretability of adversarial defense mechanisms in NIDS.

Aligned large language models (LLMs) remain vulnerable to jailbreak attacks. Recent mechanistic studies have identified latent features and representation shifts associated with jailbreak success, but they leave a more fundamental question open: why do aligned LLMs remain jailbreakable, and what structural vulnerabilities in the model make this possible? We study this question through a continuous input-transformation view. Our theoretical finding is that aligned models can still exhibit Refusal-Escape Directions (RED): local perturbation directions around a harmful input that shift the model's behavior from refusal to answering while preserving the model's harmful-semantics interpretation. From this perspective, a jailbreak is not only a successful discrete prompt construction, but can also be understood as a refusal-to-answer behavior transition induced by continuously perturbing a harmful input along RED. We then prove that RED can be exactly decomposed into contributions from operator-level sources across the model's operator structure, and identify normalization, residual-wiring, and terminal sources as analytically constrained operator-level sources. To eliminate RED, the shared expressive modules – self-attention and MLP – must eliminate the contributions from these analytically constrained sources while preserving the mechanisms that support benign responses. These competing requirements give rise to a conditional safety-utility trade-off. Experiments across multiple models and attack methods empirically analyze RED from two complementary perspectives and show that added token dimensions can expose RED, while successful jailbreaks exhibit refusal-to-answer shifts largely aligned with terminal-source contributions.

Artificial Intelligence (AI)-generated images have become increasingly realistic and readily adaptable to real-world claims, creating new challenges for verifying visual evidence. A concrete emerging risk is AI-generated refund fraud, in which manipulated or synthetic images are used to support claims about damaged products, poor delivery conditions, or service-related defects. Existing AI-generated image detection benchmarks mainly evaluate standalone authenticity classification, cross-generator transfer, or forensic localization, leaving claim-conditioned fraudulent evidence detection underexplored. To bridge this gap, we introduce FraudBench, a multimodal benchmark for detecting AI-generated fraudulent refund evidence. FraudBench is constructed from real-world user-review evidence across e-commerce, food delivery, and travel-service scenarios. We curate real evidence images together with their associated review and product metadata, identify genuine damaged and undamaged evidence through MLLM-assisted filtering and human annotation, and synthesize fake-damaged evidence from genuine undamaged reference images using 12 state-of-the-art image editing and generation models. Using FraudBench, we evaluate MLLMs, specialized AI-generated image detectors, and human evaluators on the same evaluation images. Experiments show that current MLLMs often recognize real-damaged evidence but fail on many fake-damaged subsets, with fake-damage detection rates (TPR) far below the 50% baseline on most generator subsets. Specialized detectors generally perform better but remain inconsistent across generators and can produce false positives on real-damaged samples, revealing a clear gap between generic AI image detection and reliable claim-conditioned refund-evidence verification.

Automated intrusion-style workflows require LLM agents to reason over partial observations, tool outputs, and executable artifacts under bounded budgets. A single LLM instance often compresses evidence extraction, planning, execution, and validation into one context, which increases the risk of context drift and error propagation. Existing LLM-based multi-agent systems support general collaboration, but they do not explicitly model the role boundaries, artifact provenance, and cost constraints that characterize multi-stage intrusion workflows. This paper presents CAESAR, a coordinated multi-agent framework for controlled analysis of LLM-agent behavior in intrusion-style tasks. CAESAR decomposes the workflow into five typed roles and coordinates them through a bounded round protocol with a persistent knowledge base, a per-round workspace, validator-gated knowledge promotion, and capability-token write isolation. We evaluate CAESAR on 25 CTF tasks across five categories and four LLM backends. Compared with a single-agent baseline under matched budgets and tool access, CAESAR improves task success and reduces performance variance, with larger gains on tasks requiring multi-step exploit composition. A secondary simulated interactional-security study suggests that the role structure can transfer beyond code-native surfaces. The results indicate that role transitions, artifact provenance, and knowledge-promotion events provide useful structural signals for monitoring coordinated LLM-agent behavior beyond individual prompt and output inspection. The dataset, implementation, and evaluation logs are released at https://github.com/Xu-Qiu/CMAS.

Class-level machine unlearning aims to remove the influence of specified classes while preserving model utility on retained classes. Existing methods are commonly evaluated by retain-set accuracy, forget-set accuracy, and unlearning time, but these metrics provide limited insight into how forgetting is achieved internally. In this paper, we reveal a bias-dominated shortcut in class-level unlearning: the prediction of forgotten classes can be suppressed by decreasing the corresponding bias terms in the final classification head. We first analyze the gradient dynamics of classification-head biases under softmax cross-entropy training, explaining why retain-set-only optimization tends to reduce the biases of absent classes. Based on this observation, we introduce BiasShift as a diagnostic baseline, showing that simple bias manipulation can satisfy conventional unlearning metrics while leaving abnormal bias patterns that reveal forgotten labels. To mitigate excessive forgotten-class bias suppression, we propose two bias-aware mechanisms, namely Two-Stage Bias Gradient Reversal Mechanism (TS-BGRM) and Lower-Bound Hinge Regularization (LB-HR). We further introduce three bias-oriented metrics, including Bias Stability Coefficient (BSC), Median Bias Gap (MBG), and Minimal Bias Score (MBS), to quantify bias dependence and potential leakage. Experiments on CIFAR-10, CIFAR-100, and Tiny-ImageNet demonstrate that the proposed methods maintain competitive unlearning performance while producing more stable bias distributions. We have released our code at {https://github.com/zwd2024/Beyond-the-Shadow-of-Bias-From-Classification-Head-Bias-to-Parameter-Redistribution}.

Modern cryptography is hinged on "not learning from mistakes": trying numerous wrong keys, should not help one identify the right key. Indeed, it worked – until recently when the surprising power of AI to see pattern in apparent randomness has turned the 'wrong plaintexts' generated by the 'wrong key' into productive inferential input. Crunching through these random-looking plaintext candidates AI can de-flatten the probability curve over the remaining key space. The more spiked this curve, the faster the ciphertext is defeated. This new attack vector demands a thorough review of our cryptographic security posture. NIST PQC is not immunized against AI-Accelerated Brute Force attack. Defense is rooted in non-trivial ciphertexts, in unilateral randomness, and in variable key size. This points to a new security class: Pattern Devoid Cryptography which is to be added into the toolbox used by the cyber security community.

Embedded and Internet-of-Things (IoT) devices play a critical role in modern life. Their software and firmware, often developed in memory-unsafe languages like C, are susceptible to memory safety vulnerabilities that can lead to control-flow hijacking attacks. Shadow stack is a defense mechanism against control-flow hijacking that targets return addresses. However, existing shadow stack solutions for embedded systems have the following limitations. First, they lack system-wide protection, particularly for interrupts and exceptions. Second, they introduce high performance overhead. Third, they depend on security extensions like a trusted execution environment, which are not universally available on embedded devices. Finally, they rely on hardware features that have inherent configurable constraints, which pose compatibility challenges when integrating security mechanisms that require similar hardware support. To overcome these limitations, we present WATSON, an efficient and effective shadow stack solution. It leverages a standard hardware debug unit named data watchpoints for shadow stack protection on embedded systems. To prevent unauthorized access to the shadow stack, WATSON leverages the address-matching features of the debug unit to enforce the write protection of the shadow stack. Additionally, WATSON is compatible with compiler options to enforce forward-edge control-flow integrity. We implemented a prototype of WATSON on the ARM CortexM architecture, and the concept also applies to other platforms. The introduced overhead is 7.33% and 1.81% on BEEBS and CoreMark-Pro benchmarks, respectively. We also evaluate WATSON on exception handling and two real-world applications, observing negligible performance overhead and a worst-case code size overhead of 2.11%. Furthermore, our security evaluation demonstrates that WATSON effectively prevents attacks.

This position paper argues that computer science conferences should require tamper-evident, nonrepudiable attestations of experimental results. We name the underlying problem experiment nonrepudiation: a compliant protocol must bind the numbers in a paper to an actual executed computation in a way the author cannot later alter or deny. The current system relies on self-reported checklists, optional code sharing, and author-controlled logging. None of these mechanisms answer the question a reviewer cannot check: did the code the paper describes produce the numbers the paper reports? We define the problem formally, state the security properties any compliant protocol must satisfy, and describe a threat model that includes attacks current approaches do not prevent. We frame the mechanism that provides this property as a proof-of-compute layer, a general requirement that computed results be admitted as evidence only when accompanied by a verifiable record of the computation that produced them. To show that the problem is solvable, we built K-Veritas, a reference implementation in Go that produces signed reports without accessing training data. K-Veritas is a testbed, not a finished answer. We call on conferences and the community to treat nonrepudiation as a major requirement and to help build an open, independent standard for it.

LLM-based multi-agent systems (LLM-MAS) have become a promising paradigm for solving complex tasks through role specialization, tool use, memory, and collaborative reasoning. However, these interactions create new security risks that malicious instructions injected through messages, tools, or memories can propagate across agents and rounds, causing system-level compromise. Existing defenses largely rely on local filtering or graph-based anomaly detection, but they often fail to trace fine-grained propagation paths or remediate contaminated states without disrupting benign collaboration. We propose PropGuard, a propagation-aware framework for safeguarding LLM-MAS. PropGuard constructs a dual-view spatio-temporal graph that combines response-centric risk estimation with full-state evidence preservation. Guided by these risk priors, a GE-GRPO trained inspector sequentially explores the full-state graph to recover compact suspicious propagation subgraphs. PropGuard then verifies harmful propagation through subgraph-aware diagnosis and applies source-guided remediation to correct upstream contamination and replay affected downstream interactions. Experiments across four communication architectures and five attack settings demonstrate that PropGuard consistently lowers attack success while maintaining high task-level defense success, achieving a favorable effectiveness–efficiency trade-off.

Since the official release of ChatGPT in 2022, large language models (LLMs) have rapidly evolved from chatbot-style interfaces into agentic systems that can delegate work through tools and newly spawned subagents. While these capabilities improve automation and scalability, they also pose new security risks in multi-agent networks. Existing research has studied how individual LLM-based agents can be compromised through prompt injection, jailbreaking, poisoned retrieval data, or malicious extensions. Less is known about what happens after one agent is compromised inside a multi-agent network. In particular, inherited memory from parent agents can carry malicious instructions, outdated states, or unintended behavioral rules into newly created subagents, allowing a local compromise to spread across agent boundaries. In this paper, we model contemporary multi-agent networks through the lens of subagent inheritance. Our analysis shows that current frameworks can violate trust boundaries through insecure memory inheritance, weak resource control, stale post-spawn state, and improper termination authority. We demonstrate these risks in real agent frameworks and propose defenses based on explicit security invariants. Our findings show that inheritance is not merely an implementation detail, but a central component influencing the security of multi-agent systems.

The realtime analysis and secure transmission of electrocardiogram ECG signals are critical for accurate diagnosis and safeguarding patient privacy in telemedicine applications This study presents a novel realtime ECG monitoring system that employs a learnable key generator LKG derived from each patients own ECG signal characteristics to dynamically produce unique encryption keys These keys determine the parameters r and x0 of a logistic map used for chaotic encryption The system securely encrypts realtime ECG data immediately after acquisition ensuring confidential transmission and storage in the cloud For remote clinical access the encrypted data is downloaded and decrypted on the doctors side using the matching key generated at the source or securely stored in the cloud This approach eliminates the need for traditional key exchange and substantially raises the cost of exhaustive key search in practice through persegment biometric key refresh and combined permutation and XOR diffusion supported by minentropy evaluation Compared to statickey methods the learnable biometric key design offers greater unpredictability and individualization A comprehensive set of security assessments including Shannon entropy 7678 bits correlation and autocorrelation disruption histogram statistics NIST SP 80022 frequency testing plaintextkey sensitivity avalanche effect FFTbased spectral flatness and robustness to noise and occlusion confirms the methods strength Reconstruction fidelity MSE approximately 5x106 PSNR greater than 52 dB MAE approximately 0002 demonstrates nearlossless decryption and preserved diagnostic features Encryption latency remains low preserving realtime performance.

Security Level 5 (SL5) is a security posture for AI systems that could plausibly thwart top-priority operations by the world's most cyber-capable institutions: those with extensive resources, state-level infrastructure, and expertise years ahead of the public state of the art. The SL5 terminology originates from the RAND Corporation's 2024 report "Securing AI Model Weights". Frontier AI development requires use-case-specific, productivity-optimised and updateable AI datacenter security standards. This first revision of the SL5 standard focuses on requirements with long lead times: interventions that must be planned years in advance, such as facility construction, hardware procurement, and organizational capability development. We prioritize these requirements because preserving optionality for SL5 by 2028/2029 requires starting now. These capabilities cannot be retrofitted on short notice when the need becomes urgent. Some requirements represent significant departures from current day standard practice. We believe bold measures are necessary for this level of security and see clear opportunities to apply optimization pressure to existing and novel solutions to customize them for the AI industry and address the practical operational requirements as much as possible. Our organization exists to begin paving this path. Some requirements approximate government security capabilities where private-sector approaches may be insufficient. We identify these gaps and note where government involvement may ultimately be necessary.

Federated Learning (FL) with parameter-efficient fine-tuning, such as Low-Rank Adaptation (LoRA), enables scalable model training on distributed data. However, when combined with Differential Privacy (DP), LoRA often introduces errors during global aggregation and amplifies the negative effect of DP noise. Existing cross-silo FL approaches mitigate the aggregation error by freezing one LoRA module and applying output perturbation. However, in a restricted low-rank subspaces, this additive noise frequently overwhelms the signals of the weight matrices, leading to suboptimal accuracy. To address this vulnerability, we propose FedPower, a differentially private cross-silo FL framework that reshapes server-side aggregation. Instead of perturbing mismatched low-rank factors, FedPower explicitly reconstructs and clips full-rank client updates to bound the sensitivity. The server then projects the exact aggregated update back into a secure low-rank space using PowerDP, a novel differentially private low-rank factorization mechanism. Based on simultaneous subspace iteration, PowerDP injects calibrated DP noise prior to the final orthonormalization step, effectively mitigates the negative effect of DP noise by preserving matrix orthogonality. We provide rigorous theoretical analyses establishing sensitivity bounds for subspace projections, proving that FedPower achieves both sample-level and client-level DP. Extensive experiments on various language understanding tasks in cross-silo FL settings show that FedPower is robust against tight privacy budgets while adding negligible computational overheads. Additional empirical study on different DP noise injection schemes validates the effectiveness of PowerDP in improving the tradeoff in accuracy and privacy. Evaluation on three different membership inference attacks validates the robustness and privacy-preserving capability of the proposed framework.

We discover that prompt-injection success and tool-execution success are separable safety properties: defenses that block injection do not necessarily block execution, and vice versa. We call this the injection-execution dissociation. In LLM agents with persistent memory, malicious instructions are stored at rates exceeding 97.5%, yet downstream execution ranges from 0% to 95% with no correlation to storage rate. This reframes the threat model: preventing storage alone is insufficient, and blocking execution requires structurally enforcing authority boundaries between memory ingestion and action execution. We substantiate this through a 5,040-run factorial experiment across nine open-source models (N=40 per condition), evaluating six defenses at four architectural layers against delayed-trigger attacks that persist across session boundaries via RAG retrieval. Defense effectiveness is governed by where a defense sits relative to the attack's authority boundary, not by classifier quality. Only Memory Sandbox – a tool-layer defense that structurally isolates recalled memory from executable context – reduces attack success to 0% for eight of nine models. A reasoning-mode ablation reveals a double dissociation: no single schema-layer intervention is safe across both reasoning and non-reasoning model classes. A loaded-corpus frontier evaluation (21 models, 3 providers; N=40 base, headline models topped up to N=172) reveals vendor-correlated patterns: Anthropic blocks predominantly at injection, OpenAI blocks at execution with variable generational hardening, and a pre-release Gemini endpoint exfiltrates in the majority of runs. Stored-but-dormant payloads constitute a compositional supply-chain risk in shared-memory deployments.

We present Elevator, the first binary translator that statically translates entire x86-64 executables to AArch64 without debug information, source code, or assumptions about code layout. Unlike existing systems, which rely on heuristics or runtime fallbacks to handle code-versus-data decoding errors, Elevator considers all possible interpretations of every byte and produces a separate translation for each feasible one ahead of time. Any byte may be interpreted as data, an opcode, or an opcode argument; we generate separate control flow paths for all interpretations, pruning only those leading to abnormal termination. Translations are built by composing code "tiles" automatically derived from a high-level description of the source ISA, yielding a nimble translation framework. The approach is deterministic and produces complete, self-contained binaries with no runtime component in the trusted code base. The principal cost is substantial code size expansion. The key benefit is that the output is the actual code that will run, enabling testing, validation, certification, and cryptographic signing prior to deployment, reducing risk compared to emulators or JIT compilers. We evaluate Elevator on a diverse corpus of real-world binaries, including the entire SPECint 2006 suite, demonstrating that static full-program binary translation can be both reliable and practical. Elevator achieves performance on par with or better than QEMU's user-mode JIT emulation.

Telegram has emerged as a major platform for large-scale video piracy, where copyrighted content is rapidly distributed among users. Despite its prominence, the structural and operational dynamics of this ecosystem remain insufficiently understood. To address this gap, we present the first large-scale study of video piracy on Telegram through a mixed-method analysis of 1,057 channels that shared 209k unique posts between December 2023 and January 2026 - systematically characterizing their content, distribution strategies, and how the ecosystem is sustained at scale. Central to our approach is the development of a fine-grained taxonomy that enables a structured understanding of the activity and intent of these channels on a per-post level. The channels collectively distributed 19,033 unique copyrighted titles originating from 175 countries, accumulating over 4.85B unique views and resulting in a lower-bound estimated financial loss of $17.49B for content rights holders. We also find that this ecosystem is deliberately engineered to be resilient against takedown efforts, frequently redirecting users through chains of intermediary channels and automated bots that collectively handle hosting, access control, monetization, and channel discovery. The scale and persistence of this ecosystem motivated the development of Anti-RIP, a real-time framework for detecting emerging video piracy communities on Telegram. Anti-RIP utilizes our taxonomy to generate contextual, interpretable insights that stakeholders confirmed improve the triaging action against reported posts and channels. Over a 61-day period, the framework facilitated the takedown of 524 previously unknown piracy channels and 71 bots. To support reproducibility and future research, we open-source both the dataset and the Anti-RIP framework.

While contemporary deep learning malware detectors define a dominant defense paradigm, their sophistication also exposes them to novel structural evasion attacks, a limitation we attribute to their inherent inability to express epistemic uncertainty. To address this challenge, we present MAGMA, a Retrieval-Augmented Generation (RAG) framework that decouples malware analysis into semantic code retrieval and probabilistic verification. In contrast to monolithic classifiers, MAGMA employs a dual-stream embedding scheme over assembly and pseudo-code representations to isolate Decision-Critical Functions (DCFs) from the noise of dead code. We further introduce a Stochastic Consistency Ensemble, in which multiple instances of the same reasoning agent independently evaluate the retrieval set under non-deterministic sampling. From this ensemble, we derive two complementary metrics: Function Evidence Strength (FES), a weighted aggregation of retrieval confidence, and the Evidence Conflict Score (ECS), defined as the Shannon entropy of the ensemble's predictive distribution. We show that elevated ECS values serve as an effective proxy for structural ambiguity, enabling the system to implement a principled “reject-option” policy. Extensive evaluation demonstrates that MAGMA achieves a 98.4% detection rate, substantially exceeding existing solutions.

LLM coding agents now generate code at an unprecedented scale, yet LLM-generated code introduces cybersecurity vulnerabilities into codebases without human involvement. Even when frontier models are explicitly asked to write secure production code with relevant weaknesses to avoid in context, we find that they still produce verifiable vulnerabilities on average 23% of the time across a corpus of 250 benign coding prompts. We introduce SecureForge, an automated pipeline that both audits security risks of frontier models and produces auditing-informed secure system prompts that reduce output security vulnerabilities while maintaining unit test performance. SecureForge first identifies benign prompts that produce statically detectable vulnerabilities, and then amplifies them into a large synthetic prompt corpus of diverse scenarios using a Markovian sampling technique to jointly maintain error rates and prompt diversity. This corpus is then used to iteratively optimize the system prompts to reduce output security vulnerabilities. On frontier models, SecureForge yields a statistically significant Pareto improvement in both unit test success and output security, with output vulnerabilities reduced by up to 48%. The resulting system prompts transfer zero-shot to in-the-wild coding agent prompts, without any exposure to real user prompt distributions during optimization.

Kettle is an attested build system that produces cryptographically verifiable provenance for software built inside Trusted Execution Environments (TEEs). A Kettle build records the source commit, dependency set, toolchain, build environment, and output artifact digests in a provenance document produced inside a measured confidential VM. The SHA-256 digest of that document is committed to the TEE platform's attestation report-data field, so the hardware-signed attestation report is itself the signature on the provenance, with the signing identity chaining to the TEE manufacturer's root of trust rather than to the build infrastructure operator. Because the CVM image is itself reproducible, its launch measurement is public and stable, which lets a build requester pre-attest the CVM before submitting any input and optionally deliver source over a TLS channel terminated inside it, so the build runs end-to-end confidentially without the host ever seeing source code in plaintext. Verification reduces to one signature check against the vendor root and a small set of digest comparisons, with no need to re-execute the build. The result removes the build infrastructure, its operators, and the artifact distribution channel from the trust surface a verifier must accept when deciding whether a binary corresponds to its claimed inputs.

Institutions may benefit from collaborative inference on time-series data. In settings where privacy is necessary, multi-party computation (MPC) is a straightforward approach to providing strong guarantees, yet it remains prohibitively expensive and scales poorly with modern transformer architectures. Vertical Federated Learning (VFL) offers efficiency but suffers from privacy leakage at the embedding level, and securing the entire VFL model head via MPC remains prohibitively slow and communication-heavy for larger models. To enable practical, secure inference at scale, we propose "Public/Private Hybrid Head-VFL" (PPHH-VFL). This hybrid architecture splits the model head into an efficient plaintext public head and a secure, lightweight MPC private head. By applying adversarial training to the public embeddings, we mitigate privacy leakage; concurrently, the small private head securely preserves the flow of sensitive information needed for high downstream utility. Empirical evaluations on models ranging up to 86 million parameters demonstrate that PPHH-VFL accelerates inference by up to six orders of magnitude compared to end-to-end MPC. Compared to a standard VFL+MPC baseline, our approach scales significantly better, achieving a speedup of up to 44.4x in WAN and a 91.2x reduction in communication costs (dropping from 1.7 GB to 19 MB per batch), while simultaneously improving downstream classification accuracy by 2.50% and regression RMSE by 40.7%.

Security alert screening is the downstream task of filtering, prioritizing, correlating, and contextualizing alerts for analyst attention in Security Operations Centers. This survey reviews artificial-intelligence-driven alert screening and alert-fatigue mitigation from 2015 to 2026. We synthesize 119 records, including 87 core studies, into a four-stage workflow taxonomy covering filtering, triage, correlation, and generative augmentation. We find persistent gaps in operational validation, adversarial robustness, cross-environment generalization, and evaluation practice. The survey concludes with a research agenda toward trustworthy Cognitive Security Operations Centers.

Large language models (LLMs) rely on deterministic pseudorandom number generators (PRNGs) for autoregressive sampling, creating a critical supply-chain attack surface overlooked by existing defenses. We present SeedHijack, a backdoor attack that manipulates PRNG outputs to force attacker-specified token selection without altering model logits. In a 540-trial benchmark on GPT-2 (124M), the attack achieves 99.6% exact token injection across 9 sampling configurations; it reaches 100% success on four aligned models (1.5B-7B, RLHF/SFT/reasoning distillation) and bypasses all alignment methods tested in this work. We further propose a defense based on a hardware quantum random number generator (QRNG), which neutralizes the attack in our evaluated threat model with negligible median overhead (+0.6% latency, +7.7 MB memory). Our work identifies a critical sampling-layer vulnerability and provides a practical, deployable QRNG-based defense.

Browser agents are increasingly deployed in long-horizon tasks, which require executing extended action chains to accomplish user goals. However, this prolonged execution process provides attackers with more opportunities to inject malicious instructions. Existing prompt injection attacks against browser agents expose two key gaps: (1) low effectiveness, as attacks optimized for toy baselines fail to achieve end-to-end goals in real-world scenarios with complex environments and longer steps; (2) weak stealthiness, since most attacks pit the attack goal against the user goal, causing a significant drop in system usability under attack. To address these gaps, we propose WebTrap, a mid-task hijacking injection attack. It employs multi-step instruction fusion steering to seamlessly combine both goals, enabling the agent to resume the original user task after executing the attack goal. Furthermore, we design a context-grounded generation method to align the injected content with the task environment and system instructions, maximizing the hijacking success rate. Extensive experiments on two browser agent tasks, based on extended WASP and InjecAgent environments, demonstrate that our method achieves a high attack success rate while preserving the usability of the original system. We find that WebTrap exploits the agent's navigation vulnerabilities, binding the two goals so tightly that standard defense mechanisms cannot restore the system to normal operation. These findings reveal a critical vulnerability in agent systems during long-horizon tasks that they can be stealthily hijacked.

Device-free localization trains models from heterogeneous wireless and visual sensors (e.g., Wi-Fi, LiDAR) distributed across edge devices. Federated learning offers a privacy-respecting framework, but is brittle when clients differ in sensor modality and resolution, when their data distributions drift, and when privacy noise destroys the structural signal needed for localization. We propose UMEDA, a graph federated learning framework in which clients form nodes of a global graph that share a continuous integral operator, and aggregation is reformulated as spectral signal processing on this operator. Each client encodes its local sensors with a linear-attention layer whose kernel spectrum is low-rank filtered, suppressing modality-specific residuals so clients with different sensors align in a common low-rank subspace. The server then aggregates client updates via a diffusion model over the kernel's spectral coefficients, treating updates as discretizations of a shared operator rather than topology-bound weights – this absorbs varying graph sizes and missing modalities without node-wise correspondence. To balance privacy and utility, we add an anisotropic differential-privacy mechanism that projects noise preferentially into the null space of the signal subspace, preserving dominant eigendirections while ensuring formal $(\epsilon, \delta)$-DP under gradient clipping. On MM-Fi and the RELI11D out-of-distribution benchmark, UMEDA outperforms state-of-the-art federated baselines in accuracy, convergence, and communication efficiency, particularly under high modality heterogeneity and tight privacy budgets.

GNNs have become a standard tool for learning on relational data, yet they remain highly vulnerable to backdoor attacks. Prior defenses often depend on inspecting specific subgraph patterns or node features, and thus can be circumvented by adaptive attackers. We propose PRAETORIAN, a new defense that targets intrinsic requirements of effective GNN backdoors rather than surface-level cues. Our key observation is that flipping a victim node's prediction requires substantial influence on the victim: attackers tend to either inject many trigger nodes or rely on a small set of highly influential ones. Building on this observation, PRAETORIAN (i) analyzes internal correlations within potential trigger subgraphs to detect abnormally large injected structures, and (ii) quantifies external node influence to identify triggers with disproportionate impact. Across our evaluations, PRAETORIAN reduces the average attack success rate (ASR) to 0.55% with only a 0.62% drop in clean accuracy (CA), whereas state-of-the-art defenses still yield an average ASR of >20% and a CA drop of >3% under the same conditions. Moreover, PRAETORIAN remains effective against a range of adaptive attacks, forcing adversaries to either inject many trigger nodes to achieve high ASR (>80%), which incurs a >10% CA drop, or preserve CA at the cost of limiting ASR to 18.1%. Overall, PRAETORIAN constrains attackers to an unfavorable trade-off between efficacy and detectability.

Many-shot jailbreaking (MSJ) causes safety-aligned language models to answer harmful queries by preceding them with many harmful question-answer demonstrations. We study why this attack becomes stronger as the number of demonstrations increases. Empirically, we find that MSJ induces a progressive activation drift: the representation of a fixed harmful query moves step by step away from the safety-aligned region as more harmful demonstrations are added. Theoretically, we show that this drift can be interpreted as implicit malicious fine-tuning: conditioning on N harmful demonstrations induces SGD-style updates equivalent to optimizing on the corresponding N harmful samples. This view turns the attack mechanism into a defense principle. We append a fixed one-shot safety demonstration at inference time, which induces a counteracting safety-oriented update and restores refusal behavior. The resulting method improves the model's robustness to MSJ without modifying its parameters or requiring white-box access at deployment. Code is available at https://github.com/Thecommonirin/SafeEnd.

Ensuring safe, policy-compliant outputs from large language models requires real-time content moderation that can scale across multiple safety dimensions. However, state-of-the-art guardrail models rely on autoregressive decoders with 7B–27B parameters, reformulating what is fundamentally a classification problem as sequential text generation, a design choice that incurs high latency and scales poorly to multi-aspect evaluation. In this work, we introduce GLiGuard, a 0.3B-parameter schema-conditioned bidirectional encoder adapted from GLiNER2 for LLM content moderation. The key idea is to encode task definitions and label semantics directly into the input sequence as structured token schemas, enabling simultaneous evaluation of prompt safety, response safety, refusal detection, 14 fine-grained harm categories, and 11 jailbreak strategies in a single non-autoregressive forward pass. This schema-conditioned design lets supported task and label blocks be composed directly in the input schema at inference time. Across nine established safety benchmarks, GLiGuard achieves F1 scores competitive with 7B–27B decoder-based guards despite being 23–90$\times$ smaller, while delivering up to 16$\times$ higher throughput and 17$\times$ lower latency. These results suggest that compact bidirectional encoders can approach the accuracy of much larger guard models while drastically reducing inference cost. Code and models are available at https://github.com/fastino-ai/GLiGuard.

Federated fine-tuning (FFT) has emerged as a privacy-preserving paradigm for collaboratively adapting large language models (LLMs). Built upon federated learning, FFT enables distributed agents to jointly refine a shared pretrained LLM by aggregating local LLM updates without sharing local raw data. However, FFT-based LLMs remain vulnerable to model manipulation threats, in which adversarial participants upload manipulated LLM updates that corrupt the aggregation process and degrade the performance of the global LLM. In this paper, we propose an Augmented Model maniPulation (AugMP) strategy against FFT-based LLMs. Specifically, we design a novel graph representation learning framework that captures feature correlations among benign LLM updates to guide the generation of malicious updates. To enhance manipulation effectiveness and stealthiness, we develop an iterative manipulation algorithm based on an augmented Lagrangian dual formulation. Through this formulation, malicious updates are optimized to embed adversarial objectives while preserving benign-like parameter characteristics. Experimental results across multiple LLM backbones demonstrate that the AugMP strategy achieves the strongest manipulation performance among all competing baselines, reducing the global LLM accuracy by up to \(26\%\) and degrading the average accuracy of local LLM agents by up to \(22\%\). Meanwhile, AugMP maintains high statistical and geometric consistency with benign updates, enabling it to evade conventional distance- and similarity-based defense methods.