Papers

Large language models (LLMs) are increasingly being deployed as software engineering agents that autonomously contribute to repositories. A major benefit these agents present is their ability to find and patch security vulnerabilities in the codebases they oversee. To estimate the capability of agents in this domain, we introduce ZeroDayBench, a benchmark where LLM agents find and patch 22 novel critical vulnerabilities in open-source codebases. We focus our efforts on three popular frontier agentic LLMs: GPT-5.2, Claude Sonnet 4.5, and Grok 4.1. We find that frontier LLMs are not yet capable of autonomously solving our tasks and observe some behavioral patterns that suggest how these models can be improved in the domain of proactive cyberdefense.

Micro-Controller Units (MCUs) are widely used in safety-critical systems, making them attractive targets for attacks. This calls for lightweight defenses that remain effective despite software compromise. Control Flow Auditing (CFAud) is one such mechanism wherein a remote verifier (Vrf) is guaranteed to received evidence about the control flow path taken on a prover (Prv) MCU, even when Prv software is compromised. Despite promising benefits, current CFAud architectures unfortunately require a “busy-wait” phase where a hardware-anchored root-of-trust (RoT) in Prv retains execution control to ensure delivery of control flow evidence to Vrf. This drastically reduces the CPU utilization on Prv. In this work, we addresses this limitation with an architecture for Contention Avoidance in Runtime Auditing with Minimized Execution Latency (CARAMEL). CARAMEL is a hardware-software RoT co-design that enables Prv applications to resume while control flow evidence is transmitted to Vrf. This significantly reduces contention due to transmission delays and improves CPU utilization without giving up on security. Key to CARAMEL is our design of a new RoT with a self-contained (and minimal) dedicated communication interface. CARAMEL's implementation and accompanying evaluation are made open-source. Our results show substantially improved CPU utilization at a modest hardware cost.

We consider an asynchronous network of $n$ parties connected to each other via secure channels, up to $t$ of which are byzantine. We study common coin tossing, a task where the parties try to agree on an unpredictable random value, with some chance of failure due to the byzantine parties' influence. Coin tossing is a well-known and often-studied task due to its use in byzantine agreement. In this work, we present a committee-based method to transform strong (rarely failing) binary common coins into weaker ones that asymptotically require less communication. For any $k > 2$ and $\varepsilon > 0$, we can transform a strong binary coin that costs $\widetilde{O}(n^k)$ bits of communication into a weak binary coin that costs $\widetilde{O}(\varepsilon^{-2k}n^{3 - 2/k})$ bits. This latter coin tolerates $\varepsilon n$ fewer byzantine parties than the strong coin it is based on, and it fails with an arbitrarily small constant probability. With our method, we obtain a secure-channel-based perfectly secure coin for $t \leq (\frac{1}{4} - \varepsilon)n$ faults that costs $\widetilde{O}(n^{2.5})$ bits, as well as a coin based on cryptographic hashing for $t \leq (\frac{1}{3} - \varepsilon)n$ faults that costs $\widetilde{O}(n^{7/3}\kappa)$ bits. These are to our knowledge the first PKI-free asynchronous common coins that cost $o(n^3)$ bits of communication but still succeed with at least constant probability against $t = \Theta(n)$ adaptive byzantine faults.

Federated Learning (FL) was initially proposed as a privacy-preserving machine learning paradigm. However, FL has been shown to be susceptible to a series of privacy attacks. Recently, there has been concern about the Source Inference Attack (SIA), where an honest-but-curious central server attempts to identify exactly which client owns a given data point which was used in the training phase. Alarmingly, standard gradient obfuscation techniques with Differential Privacy have been shown to be ineffective against SIAs, at least without severely diminishing the accuracy. In this work, we propose a defense against SIAs within the widely studied shuffle model of FL, where an honest shuffler acts as an intermediary between the clients and the server. First, we demonstrate that standard naive shuffling alone is insufficient to prevent SIAs. To effectively defend against SIAs, shuffling needs to be applied at a more granular level; we propose a novel combination of parameter-level shuffling with the residue number system (RNS). Our approach provides robust protection against SIAs without affecting the accuracy of the joint model and can be seamlessly integrated into other privacy protection mechanisms. We conduct experiments on a series of models and datasets, confirming that standard shuffling approaches fail to prevent SIAs and that, in contrast, our proposed method reduce the attack's accuracy to the level of random guessing.

We study the problem of computing a U-statistic with a kernel function f of degree k $\ge$ 2, i.e., the average of some function f over all k-tuples of instances, in a federated learning setting. Ustatistics of degree 2 include several useful statistics such as Kendall's $\tau$ coefficient, the Area under the Receiver-Operator Curve and the Gini mean difference. Existing methods provide solutions only under the lower-utility local differential privacy model and/or scale poorly in the size of the domain discretization. In this work, we propose a protocol that securely computes U-statistics of degree k $\ge$ 2 under central differential privacy by leveraging Multi Party Computation (MPC). Our method substantially improves accuracy when compared to prior solutions. We provide a detailed theoretical analysis of its accuracy, communication and computational properties. We evaluate its performance empirically, obtaining favorable results, e.g., for Kendall's $\tau$ coefficient, our approach reduces the Mean Squared Error by up to four orders of magnitude over existing baselines.

Every year, millions of functional systems become e-waste because users are pressured to send their systems to landfills due to a lack of vendor support and difficulty in recycling. Vendors cite “cybersecurity” as the driver for short product support periods, leading to a prevalent, but uninterrogated, belief that cybersecurity and environmental sustainability are fundamentally contradictory; i.e., it is difficult, if not impossible, to build products that are secure, long-lasting, and reusable. To understand the nuanced relationship between security and sustainability, we systematically analyze 29 papers and distill 155 sustainability guidelines into 12 sustainability themes. These themes enable us to compare the sustainable HCI and sustainable software engineering guidance with that of cybersecurity, identifying points of alignment and tension. We find little evidence of a fundamental tension between these two domains; the few instances of tension can be mitigated through thoughtful consideration of security and sustainability objectives. We also find that sustainability, like usable security, struggles with the myth of users as the weakest link and the individualization of responsibility. Building on these parallels, we argue that the usable security community is well-positioned to integrate sustainability considerations, as both fields share challenges in shifting responsibility from individuals to systemic design.

Access to frontier large language models (LLMs), such as GPT-5 and Gemini-2.5, is often hindered by high pricing, payment barriers, and regional restrictions. These limitations drive the proliferation of $\textit{shadow APIs}$, third-party services that claim to provide access to official model services without regional limitations via indirect access. Despite their widespread use, it remains unclear whether shadow APIs deliver outputs consistent with those of the official APIs, raising concerns about the reliability of downstream applications and the validity of research findings that depend on them. In this paper, we present the first systematic audit between official LLM APIs and corresponding shadow APIs. We first identify 17 shadow APIs that have been utilized in 187 academic papers, with the most popular one reaching more than 5,900 citations and 58,000 GitHub stars by December 6, 2025. Through multidimensional auditing of three representative shadow APIs across utility, safety, and model verification, we uncover widespread behavioral inconsistency and fingerprint-based evidence consistent with deceptive model claims in a subset of audited endpoints. Specifically, we reveal performance divergence reaching up to 47.21%, significant unpredictability in safety behaviors, and identity verification failures in 45.83% of fingerprint tests. These practices critically undermine the reproducibility and validity of scientific research, harm the interests of shadow API users, and damage the reputation of official model providers. By the time of writing, 4 of the 17 providers have already ceased operations, underscoring the operational volatility of this market. Meanwhile, unverifiable compliance claims and independent model-substitution testing platforms have emerged in the ecosystem, reflecting growing community awareness of this risk.

The proliferation of IoT and V2X systems generates unprecedented sensitive data at the network edge, demanding privacy-preserving architectures that enable secure sharing without exposing raw information. Contemporary solutions face a fundamental privacy-efficiency-trust trilemma: achieving strong privacy guarantees, computational efficiency for resource-constrained devices, and decentralized trust simultaneously remains intractable with single-paradigm approaches. This survey systematically analyzes 75 technical papers (2007–2025) through a novel three-dimensional taxonomy classifying architectures into Decentralized Computation, Cryptography-based, and Distributed Ledger approaches. Temporal analysis reveals dramatic acceleration during 2024–2025, with 48% of all papers published in this period – Decentralized Computation dominates at 44% of contributions and 59% of 2025 publications. Comprehensive Security Threat Mapping and Technology Maturity Assessment demonstrate that mature solutions occupy narrow design regions excelling in one or two dimensions while compromising others, conclusively validating the trilemma hypothesis. We identify emerging hybrid architectures combining complementary paradigms as the essential path forward. Critical challenges including security guarantee composition across layers, multi-layer coordination overhead minimization, and post-quantum security integration must be addressed for practical deployment in next-generation intelligent transportation systems and IoT ecosystems.

Phishing remains the most pervasive threat to the Web, enabling large-scale credential theft and financial fraud through deceptive webpages. While recent reference-based and generative-AI-driven phishing detectors achieve strong accuracy, their reliance on external knowledge bases, cloud services, and complex multimodal pipelines fundamentally limits practicality, scalability, and reproducibility. In contrast, conventional deep learning approaches often fail to generalize to evolving phishing campaigns. We introduce SpecularNet, a novel lightweight framework for reference-free web phishing detection that demonstrates how carefully designed compact architectures can rival heavyweight systems. SpecularNet operates solely on the domain name and HTML structure, modeling the Document Object Model (DOM) as a tree and leveraging a hierarchical graph autoencoding architecture with directional, level-wise message passing. This design captures higher-order structural invariants of phishing webpages while enabling fast, end-to-end inference on standard CPUs. Extensive evaluation against 13 state of the art phishing detectors, including leading reference-based systems, shows that SpecularNet achieves competitive detection performance with dramatically lower computational cost. On benchmark datasets, it reaches an F1 score of 93.9%, trailing the best reference-based method slightly while reducing inference time from several seconds to approximately 20 milliseconds per webpage. Field and robustness evaluations further validate SpecularNet in real-world deployments, on a newly collected 2026 open-world dataset, and against adversarial attacks.

This is the Replicated Computational Results (RCR) Report for the paper “Can LLMs Hack Enterprise Networks?" The paper empirically investigates the efficacy and effectiveness of different LLMs for penetration-testing enterprise networks, i.e., Microsoft Active Directory Assumed-Breach Simulations. This RCR report describes the artifacts used in the paper, how to create an evaluation setup, and highlights the analysis scripts provided within our prototype.

Adversarial behavior plays a central role in aligning large language models with human values. However, existing alignment methods largely rely on static adversarial settings, which fundamentally limit robustness, particularly in multimodal settings with a larger attack surface. In this work, we move beyond static adversarial supervision and introduce co-evolutionary alignment with evolving attacks, instantiated by CEMMA (Co-Evolutionary Multi-Modal Alignment), an automated and adaptive framework for multimodal safety alignment. We introduce an Evolutionary Attacker that decomposes adversarial prompts into method templates and harmful intents. By employing genetic operators, including mutation, crossover, and differential evolution, it enables simple seed attacks to inherit the structural efficacy of sophisticated jailbreaks. The Adaptive Defender is iteratively updated on the synthesized hard negatives, forming a closed-loop process that adapts alignment to evolving attacks. Experiments show that the Evolutionary Attacker substantially increases red-teaming jailbreak attack success rate (ASR), while the Adaptive Defender improves robustness and generalization across benchmarks with higher data efficiency, without inducing excessive benign refusal, and remains compatible with inference-time defenses such as AdaShield.

Digital twins (DTs) are increasingly used to monitor and secure Industrial Control Systems (ICS), yet detecting stealthy False Data Injection Attacks (FDIAs) that manipulate system states within normal physical bounds remains challenging. Deep learning anomaly detectors often over-generalize such subtle manipulations, while classical fault detection methods do not scale well in highly correlated multivariate systems. We propose a closed-loop Information-Theoretic Digital Twin (IT-DT) framework for real-time anomaly detection. N4SID identification is combined with steady-state Kalman filtering to quantify residual distribution shifts via closed-form KL divergence, capturing both mean deviations and malicious cross-covariance shifts. Evaluations on the SWaT and WADI datasets show that IT-DT achieves F1-scores of 0.832 and 0.615, respectively, with better precision than deep learning baselines such as TranAD. Computational profiling indicates that the analytical approach requires minimal memory and provides approximately a 600x inference speedup over transformer-based methods on CPU hardware. This makes the framework suitable for resource-constrained industrial edge controllers without GPU acceleration.

Recent intelligent systems integrate powerful Large Language Models (LLMs) through APIs, but their trustworthiness may be critically undermined by targeted attacks like backdoor and prompt injection attacks, which secretly force LLMs to generate specific malicious sequences. Existing defensive approaches for such threats typically rely on high access rights, impose prohibitive costs, and hinder normal inference, rendering them impractical for real-world scenarios. To solve these limitations, we introduce DualSentinel, a lightweight and unified defense framework that can accurately and promptly detect the activation of targeted attacks alongside the LLM generation process. We first identify a characteristic of compromised LLMs, termed Entropy Lull: when a targeted attack successfully hijacks the generation process, the LLM exhibits a distinct period of abnormally low and stable token probability entropy, indicating it is following a fixed path rather than making creative choices. DualSentinel leverages this pattern by developing an innovative dual-check approach. It first employs a magnitude and trend-aware monitoring method to proactively and sensitively flag an entropy lull pattern at runtime. Upon such flagging, it triggers a lightweight yet powerful secondary verification based on task-flipping. An attack is confirmed only if the entropy lull pattern persists across both the original and the flipped task, proving that the LLM's output is coercively controlled. Extensive evaluations show that DualSentinel is both highly effective (superior detection accuracy with near-zero false positives) and remarkably efficient (negligible additional cost), offering a truly practical path toward securing deployed LLMs. The source code can be accessed at https://doi.org/10.5281/zenodo.18479273.

Large Language Models (LLMs) are increasingly deployed as agentic systems that plan, memorize, and act in open-world environments. This shift brings new security problems: failures are no longer only unsafe text generation, but can become real harm through tool use, persistent memory, and interaction with untrusted web content. In this survey, we provide a transition-oriented view from Secure Agentic AI to a Secure Agentic Web. We first summarize a component-aligned threat taxonomy covering prompt abuse, environment injection, memory attacks, toolchain abuse, model tampering, and agent network attacks. We then review defense strategies, including prompt hardening, safety-aware decoding, privilege control for tools and APIs, runtime monitoring, continuous red-teaming, and protocol-level security mechanisms. We further discuss how these threats and mitigations escalate in the Agentic Web, where delegation chains, cross-domain interactions, and protocol-mediated ecosystems amplify risks via propagation and composition. Finally, we highlight open challenges for web-scale deployment, such as interoperable identity and authorization, provenance and traceability, ecosystem-level response, and scalable evaluation under adaptive adversaries. Our goal is to connect recent empirical findings with system-level requirements, and to outline practical research directions toward trustworthy agent ecosystems.

Compliance as code is an emerging idea about automating compliance through programmed compliance controls and checks. Given scant existing research thus far, the paper presents an empirical analysis of a compliance as code project addressing open source software (OSS) projects and products. The dataset examined covers a little over 1,500 unique compliance rules designed and implemented for 14 Linux distribution releases from five vendors. According to the results, (1) the coverage of the rules varies across the five vendors. Then, (2) the brief rationales provided for the rules do not exhibit statistical similarities but the short code snippets for these do show similarities to some extent. Furthermore, (3) as many as 24 controls are covered from over 10 different organizations, among them governmental agencies, standardization organizations, and non-profit associations. Finally, (4) the rules can be mapped to the essential cyber security requirements of the Cyber Resilience Act (CRA), although only modest agreement exists among the three authors regarding individual mappings. This observation supports an argument that the compliance as code project studied could be updated with new compliance checks. Given that also operating systems are in the CRA's scope when used in a network-connected product, such an updating would have also practical relevance in the nearby future.

The rapid development of large language models (LLMs) has driven the widespread adoption of cloud-based LLM inference services, while also bringing prominent privacy risks associated with the transmission and processing of private data in remote inference. For privacy-preserving LLM inference technologies to be practically applied in industrial scenarios, three core requirements must be satisfied simultaneously: (1) Accuracy and efficiency losses should be minimized to mitigate degradation in service experience. (2) The inference process can be run on large-scale clusters consist of heterogeneous legacy xPUs. (3) Compatibility with existing LLM infrastructures should be ensured to reuse their engineering optimizations. To the best of our knowledge, none of the existing privacy-preserving LLM inference methods satisfy all the above constraints while delivering meaningful privacy guarantees. In this paper, we propose AloePri, the first privacy-preserving LLM inference method for industrial applications. AloePri protects both the input and output data by covariant obfuscation, which jointly transforms data and model parameters to achieve better accuracy and privacy. We carefully design the transformation for each model component to ensure inference accuracy and data privacy while keeping full compatibility with existing infrastructures of Language Model as a Service. AloePri has been integrated into an industrial system for the evaluation of mainstream LLMs. The evaluation on Deepseek-V3.1-Terminus model (671B parameters) demonstrates that AloePri causes accuracy loss of 0.0%~3.5% and exhibits efficiency equivalent to that of plaintext inference. Meanwhile, AloePri successfully resists state-of-the-art attacks, with less than 5% of tokens recovered. To the best of our knowledge, AloePri is the first method to exhibit practical applicability to large-scale models in real-world systems.

Large Language Models (LLMs) are increasingly deployed for code generation in high-stakes software development, yet their limited transparency in security reasoning and brittleness to evolving vulnerability patterns raise critical trustworthiness concerns. Models trained on static datasets cannot readily adapt to newly discovered vulnerabilities or changing security standards without retraining, leading to the repeated generation of unsafe code. We present a principled approach to trustworthy code generation by design that operates as an inference-time safety mechanism. Our approach employs retrieval-augmented generation to surface relevant security risks in generated code and retrieve related security discussions from a curated Stack Overflow knowledge base, which are then used to guide an LLM during code revision. This design emphasizes three aspects relevant to trustworthiness: (1) interpretability, through transparent safety interventions grounded in expert community explanations; (2) robustness, by allowing adaptation to evolving security practices without model retraining; and (3) safety alignment, through real-time intervention before unsafe code reaches deployment. Across real-world and benchmark datasets, our approach improves the security of LLM-generated code compared to prompting alone, while introducing no new vulnerabilities as measured by static analysis. These results suggest that principled, retrieval-augmented inference-time interventions can serve as a complementary mechanism for improving the safety of LLM-based code generation, and highlight the ongoing value of community knowledge in supporting trustworthy AI deployment.

Hotaru Beam is a logic puzzle which objective is to connect circles placed on a grid by drawing only lines with specified starting points and numbers of bends. A zero-knowledge proof is a communication protocol that allows one player to persuade the other that they are in possession of a certain piece of information without actually revealing it. We show that Hotaru Beam is NP-complete and present a physical zero-knowledge proof (i.e. implementable using physical items) for proving that one knows a solution to the puzzle.

Automating network security analysis, particularly the identification of potential attack paths, presents significant challenges. Due in part to the sequential, interconnected, and evolutionary nature of system events which most artificial intelligence (AI) techniques struggle to model effectively. This paper proposes a Reinforcement Learning (RL) environment generation framework that simulates the sequence of processes executed on a Windows operating system, enabling dynamic modeling of malicious processes on a system. This methodology models operating system state and transitions using a graph representation. This graph is derived from open-source System Monitor (Sysmon) logs. To address the variety in system event types, fields, and log formats, a mechanism was developed to capture and model parent-child processes from Sysmon logs. A Gymnasium environment (SubstratumGraphEnv) was constructed to establish the perceptible basis for an RL environment, and a customized PyTorch interface was also built (SubstratumBridge) to translate Gymnasium graphs into Deep Reinforcement Learning (DRL) observations and discrete actions. Graph Convolutional Networks (GCNs) concretize the graph's local and global state, which feed the distinct policy and critic heads of an Advantage Actor-Critic (A2C) model. This work's central contribution lies in the design of a novel deep graphical RL environment that automates translation of sequential user and system events, furnishing crucial context for cybersecurity analysis. This work provides a foundation for future research into shaping training parameters and advanced reward shaping, while also offering insight into which system events attributes are critical to training autonomous RL agents.

Large language models (LLMs) increasingly act as autonomous agents, using tools to execute code, read and write files, and access networks, creating novel security risks. To mitigate these risks, agents are commonly deployed and evaluated in isolated "sandbox" environments, often implemented using Docker/OCI containers. We introduce SANDBOXESCAPEBENCH, an open benchmark that safely measures an LLM's capacity to break out of these sandboxes. The benchmark is implemented as an Inspect AI Capture the Flag (CTF) evaluation utilising a nested sandbox architecture with the outer layer containing the flag and no known vulnerabilities. Following a threat model of a motivated adversarial agent with shell access inside a container, SANDBOXESCAPEBENCH covers a spectrum of sandboxescape mechanisms spanning misconfiguration, privilege allocation mistakes, kernel flaws, and runtime/orchestration weaknesses. We find that, when vulnerabilities are added, LLMs are able to identify and exploit them, showing that use of evaluation like SANDBOXESCAPEBENCH is needed to ensure sandboxing continues to provide the encapsulation needed for highly-capable models.

In this paper, we analyze the finality of the Filecoin network, focusing on dynamic probabilistic guarantees of tipset permanence in the canonical chain. Our approach differs from static analyses that consider only the worst-case scenario; instead, we dynamically compute the error probability at each round using the live chain history, providing a more accurate and efficient assessment. We provide a practical algorithm that only requires visibility into the blocks produced by honest participants, which can be implemented by clients or off-chain applications without any change to Filecoin's consensus mechanisms.We demonstrate that, under typical operating conditions, the sought-after error probability of $2^{-30}$ is achievable in approximately 30 rounds, a 30x improvement over the 900 rounds that the network currently encodes as a fixed threshold. This finding immediately expedites transactions and enhances usability of the Filecoin network, while laying the foundation for further analysis of other DAG-structured blockchains.

Large language models (LLMs) have shown promise for automated patching, but their effectiveness depends strongly on how they are integrated into patching systems. While prior work explores prompting strategies and individual agent designs, the field lacks a systematic comparison of patching architectures. In this paper, we present a controlled evaluation of four LLM-based patching paradigms – fixed workflow, single-agent system, multi-agent system, and general-purpose code agents – using a unified benchmark and evaluation framework. We analyze patch correctness, failure modes, token usage, and execution time across real-world vulnerability tasks. Our results reveal clear architectural trade-offs: fixed workflows are efficient but brittle, single-agent systems balance flexibility and cost, and multi-agent designs improve generalization at the expense of substantially higher overhead and increased risk of reasoning drift on complex tasks. Surprisingly, general-purpose code agents achieve the strongest overall patching performance, benefiting from general-purpose tool interfaces that support effective adaptation across vulnerability types. Overall, we show that architectural design and iteration depth, rather than model capability alone, dominate the reliability and cost of LLM-based automated patching.

Safety alignment in large language models (LLMs), particularly for cybersecurity tasks, primarily focuses on preventing misuse. While this approach reduces direct harm, it obscures a complementary failure mode: denial of assistance to legitimate defenders. We study Defensive Refusal Bias – the tendency of safety-tuned frontier LLMs to refuse assistance for authorized defensive cybersecurity tasks when those tasks include similar language to an offensive cyber task. Based on 2,390 real-world examples from the National Collegiate Cyber Defense Competition (NCCDC), we find that LLMs refuse defensive requests containing security-sensitive keywords at $2.72\times$ the rate of semantically equivalent neutral requests ($p < 0.001$). The highest refusal rates occur in the most operationally critical tasks: system hardening (43.8%) and malware analysis (34.3%). Interestingly, explicit authorization, where the user directly instructs the model that they have authority to complete the target task, increases refusal rates, suggesting models interpret justifications as adversarial rather than exculpatory. These findings are urgent for interactive use and critical for autonomous defensive agents, which cannot rephrase refused queries or retry. Our findings suggest that current LLM cybersecurity alignment relies on semantic similarity to harmful content rather than reasoning about intent or authorization. We call for mitigations that analyze intent to maximize defensive capabilities while still preventing harmful compliance.

Fine-tuning large language models (LLMs) on custom datasets has become a standard approach for adapting these models to specific domains and applications. However, recent studies have shown that such fine-tuning can lead to significant degradation in the model's safety. Existing defense methods operate at the sample level and often suffer from an unsatisfactory trade-off between safety and utility. To address this limitation, we perform a systematic token-level diagnosis of safety degradation during fine-tuning. Based on this, we propose token-level data selection for safe LLM fine-tuning (TOSS), a novel framework that quantifies the safety risk of each token by measuring the loss difference between a safety-degraded model and a utility-oriented model. This token-level granularity enables accurate identification and removal of unsafe tokens, thereby preserving valuable task-specific information. In addition, we introduce a progressive refinement strategy, TOSS-Pro, which iteratively enhances the safety-degraded model's ability to identify unsafe tokens. Extensive experiments demonstrate that our approach robustly safeguards LLMs during fine-tuning while achieving superior downstream task performance, significantly outperforming existing sample-level defense methods. Our code is available at https://github.com/Polly-LYP/TOSS.

The rapid proliferation of autonomous AI agents is driving a shift toward agentic commerce, where agents are expected to autonomously invoke and pay for services. While blockchain-based payments offer a programmable foundation for such interactions, the recently proposed x402 standard fails to enforce end-to-end atomicity across service execution, payment, and result delivery. In this paper, we present A402, a trust-minimized payment architecture that securely binds cryptocurrency payments to service execution. A402 introduces Atomic Service Channels (ASCs), a new channel protocol that integrates service execution into payment channels, enabling real-time, high-frequency micropayments for agentic commerce. Within each ASC, A402 employs an atomic exchange protocol based on TEE-assisted adaptor signatures, ensuring that payments are finalized if and only if the requested service is correctly executed and the corresponding result is delivered. To further ensure privacy, A402 incorporates a TEE-based Liquidity Vault that privately manages the lifecycle of ASCs and aggregates their settlements into a single on-chain transaction, revealing only aggregated balances. We implement A402 and evaluate it against x402 with integrations on both Bitcoin and Ethereum. Our results show that A402 delivers orders-of-magnitude performance and on-chain cost improvements over x402 while providing trust-minimized security guarantees.

An Adaptive Cruise Control (ACC) system automatically adjusts the host vehicle's speed to maintain a safe following distance from a lead vehicle. In typical implementations, a feedback controller (e.g., a Proportional-Integral-Derivative (PID) controller) computes the host vehicle's acceleration using a target speed and a spacing error, defined as the difference between the measured inter-vehicle distance and a desired safe distance. ACC is often assumed to be resilient to fault-injection attacks because a Kalman filter (KF) can smooth noisy speed measurements. However, we show–through analytical proofs and simulation results–that a KF can tolerate injected speed values only up to a bounded threshold. When injected values exceed this threshold, the filter can be driven off track, causing the ACC controller to make unsafe acceleration decisions and potentially leading to collisions. Our main contribution is to augment the PID-based controller with Intrusion Detection System (IDS) outputs, yielding Intrusion Detection Systems-Based Adaptive Cruise Control (ACC-IDS). The ACC-IDS controller is simple and implementable: a binary intrusion flag switches the control law to emergency braking. We prove that augmenting ACC with an IDS, under assumed detection-performance and latency constraints, can mitigate these attacks and help preserve ACC's collision-avoidance guarantees.

This work presents ATLAS, an LLM-driven framework that bridges standardized threat modeling and property-based formal verification for System-on-Chip (SoC) security. Starting from vulnerability knowledge bases such as Common Weakness Enumeration (CWE), ATLAS identifies SoC-specific assets, maps relevant weaknesses, and generates assertion-based security properties and JasperGold scripts for verification. By combining asset-centric analysis with standardized threat model templates and multi-source SoC context, ATLAS automates the transformation from vulnerability reasoning to formal proof. Evaluated on three HACK@DAC benchmarks, ATLAS detected 39/48 CWEs and generated correct properties for 33 of those bugs, advancing automated, knowledge-driven SoC security verification toward a secure-by-design paradigm.

Static Application Security Testing (SAST) tools often suffer from high false positive rates, leading to alert fatigue that consumes valuable auditing resources. Recent efforts leveraging Large Language Models (LLMs) as filters offer limited improvements; however, these methods treat LLMs as passive, stateless classifiers, which lack project-wide context and the ability to learn from analyses to discover unknown, similar vulnerabilities. In this paper, we propose vEcho, a novel framework that transforms the LLM from a passive filter into a virtual security expert capable of learning, memory, and reasoning. vEcho equips its core reasoning engine with a robust developer tool suite for deep, context-aware verification. More importantly, we introduce a novel Echoic Vulnerability Propagation (EVP) mechanism. Driven by a Cognitive Memory Module that simulates human learning, EVP enables vEcho to learn from verified vulnerabilities and proactively infer unknown, analogous flaws, achieving a paradigm shift from passive verification to active discovery. Extensive experiments on the CWE-Bench-Java dataset demonstrate vEcho's dual advantages over the state-of-the-art baseline, IRIS. Specifically, vEcho achieves a 65% detection rate, marking a 41.8% relative improvement over IRIS's 45.83%. Crucially, it simultaneously addresses alert fatigue by reducing the false positive rate to 59.78%, a 28.3% relative reduction from IRIS's 84.82%. Furthermore, vEcho proactively identified 37 additional known vulnerabilities beyond the 120 documented in the dataset, and has discovered 51 novel 0-day vulnerabilities in open-source projects.

Harvest-now, decrypt-later (HN-DL) attacks threaten today's encrypted communications by archiving ciphertext until a quantum computer can break the underlying key exchange. This paper reframes HN-DL as an economic problem, quantifying adversary costs across Transport Layer Security (TLS) 1.2, TLS 1.3, QUIC, and Secure Shell (SSH) with an open-source testbed that reproduces the full attack sequence. Our model shows that retaining intercepted traffic is economically trivial, shifting the defensive question from whether an adversary can archive to how much decryption will cost. We evaluate protocol configuration strategies that act along two independent cost axes: storage overhead and quantum workload. Beyond the ongoing migration to post-quantum cryptography, these strategies provide defense in depth with current infrastructure. Encrypted Client Hello forces indiscriminate bulk collection, inflating the archive the adversary must retain, while aggressive rekeying and larger key exchange parameters multiply the quantum computations required to recover plaintext. Because storage inflation penalizes both sides while quantum cost inflation targets the adversary alone, rekeying and key size selection offer the strongest defensive levers.

Watermarking has emerged as a key defense against the misuse of machine-generated images (MGIs). Yet the robustness of these protections remains underexplored. To reveal the limits of SOTA proactive image watermarking defenses, we propose HIDE&SEEK (HS), a suite of versatile and cost-effective attacks that reliably remove embedded watermarks while preserving high visual fidelity.