Papers

Existing high performance blockchains verify one signature per transaction on the critical path, which creates O(N) verification cost, high hardware pressure, and difficult post quantum migration. This paper presents ACE Runtime, a ZKP native execution layer built on identity authorization separation. We replace per transaction signature checks with lightweight HMAC attestations in the hot path, then generate one aggregated zero knowledge finality certificate per block in an asynchronous prove stage. The system is organized as an Attest Execute Prove pipeline with two tier finality: soft finality from BFT voting and hard finality from proof verification. Under standard cryptographic assumptions, we provide formal arguments for attestation unforgeability and hard finality irreversibility. We also define a two phase timeout and backup proving path with witness availability gossip for liveness under builder failure. Quantitative results combine analytical modeling with reference implementation measurements. The prototype shows low CPU orchestration overhead, while model driven analysis projects constant per block verification cost, lower validator hardware requirements for non builders, and better bandwidth efficiency than per transaction signature designs. These results indicate that identity authorization separation is a practical architecture for sub second cryptographic finality with a clear path toward stronger post quantum components.

Enterprises and organizations today increasingly deploy in-house, cloud based applications and APIs for internal operations or external customers. These deployments deal with increasing number of threats, despite security features offered by cloud service providers. This work focus on threats that exploit application layer vulnerabilities of cloud workloads. Prevention and mitigation measures against such threats need to be cognizant of application semantics, posing a hurdle to existing solutions. In this work, we design and implement a security framework that allow cloud workload administrators to easily define and enforce policies capable of preventing (i) unrestricted resource consumption, (ii) unrestricted access to sensitive business flows, and (iii) broken authentication. Our framework, Paladin, leverages large language models to extract sufficient semantic meaning from API requests to provide cloud administrators with an application agnostic policy definition interface. Once defined, requests are automatically matched with relevant policies and enforced by high performance proxies. Evaluations with our prototype show that such a framework has broad applicability across applications, good policy identification accuracy, and reasonable overheads, making it substantially easier to define and enforce cross application policies.

Considering the rise of cyberattacks incidents worldwide, the need to ensure stronger passwords is necessary. Developing a password strength meter (PSM) can help users create stronger passwords when creating an account on an online platform. This research aimed to explore whether incorporating a non-English training dataset (specifically Indian) can improve the performance of a PSM. Findings show that PSMs can be improved by utilising learning of words from other languages. Another contribution of the research was to compare and provide an analysis of AI generated data (specifically by ChatGPT) and PassGAN (existing state-of-the-art model), proving that PassGAN-like tools may no longer be needed as the performance is higher using AI generated data. To further strengthen detection, a Jaro similarity-based matching mechanism was incorporated, enabling the classification of passwords that are highly similar to known weak passwords - this addresses limitations of direct matching techniques used in prior work. A final novel contribution is on developing a PSM tailored for Indian passwords, which has not been developed previously - this resulted in a near-perfect matching accuracy using a Jaro function value of 0.5. Although performance improvements were constrained by limited data and training, results suggest that using the ChatGPT dataset is a viable and effective strategy for developing secure, language-aware password strength meters.

Model Context Protocol (MCP) servers have rapidly emerged over the past year as a widely adopted way to enable Large Language Model (LLM) agents to access dynamic, real-world tools. As MCP servers proliferate and become easy to adopt via open-source releases, understanding their security risks becomes essential for dependable production agent deployments. Recent work has developed MCP threat taxonomies, proposed mitigations, and demonstrated practical attacks. However, to the best of our knowledge, no prior study has conducted a systematic, large-scale assessment of weaknesses in open-source MCP servers. Motivated by this gap, we apply static code analysis to identify Common Weakness Enumeration (CWE) weaknesses and map them to common attack patterns and threat categories using the MITRE Common Attack Pattern Enumerations and Classifications (CAPEC) to ground risk in real-world threats. We then introduce a risk-assessment framework for the MCP landscape that combines these threats using a multi-metric scoring of likelihood and impact. Our findings show that many open-source MCP servers contain exploitable weaknesses that can compromise confidentiality, integrity, and availability, underscoring the need for secure-by-design MCP server development.

The Model Context Protocol (MCP) is a recently proposed interoperability standard that unifies how AI agents connect with external tools and data sources. By defining a set of common client-server message exchange clauses, MCP replaces fragmented integrations with a standardized, plug-and-play framework. However, to be compatible with diverse AI agents, the MCP specification relaxes many behavioral constraints into optional clauses, leading to misuse-prone SDK implementation. We identify it as a new attack surface that allows adversaries to achieve multiple attacks (e.g, silent prompt injection, DoS, etc.), named as compatibility-abusing attacks. In this work, we present the first systematic framework for analyzing this new attack surface across multi-language MCP SDKs. First, we construct a universal and language-agnostic intermediate representation (IR) generator that normalizes SDKs of different languages. Next, based on the new IR, we propose auditable static analysis with LLM-guided semantic reasoning for cross-language/clause compliance analysis. Third, by formalizing the attack semantics of the MCP clauses, we build three attack modalities and develop a modality-guided pipeline to uncover exploitable non-compliance issues.

The US Census Bureau Disclosure Avoidance System (DAS) balances confidentiality and utility requirements for the decennial US Census (Abowd et al., 2022). The DAS was used in the 2020 Census to produce demographic datasets critically used for legislative apportionment and redistricting, federal and state funding allocation, municipal and infrastructure planning, and scientific research. At the heart of DAS is TopDown, a heuristic post-processing method that combines billions of private noisy measurements across six geographic levels in order to produce new estimates that are consistent, more accurate, and satisfy certain structural constraints on the data. In this work, we introduce BlueDown, a new post-processing method that produces more accurate, consistent estimates while satisfying the same privacy guarantees and structural constraints. We obtain especially large accuracy improvements for aggregates at the county and tract levels on evaluation metrics proposed by the US Census Bureau. From a technical perspective, we develop a new algorithm for generalized least-squares regression that leverages the hierarchical structure of the measurements and that is statistically optimal among linear unbiased estimators. This reduces the computational dependence on the number of geographic regions measured from matrix multiplication time, which would be infeasible for census-scale data, to linear time. We incorporate the additional structural constraints by combining this regression algorithm with an optimization routine that extends TDA to support correlated measurements. We further improve the efficiency of our algorithm using succinct linear-algebraic operations that exploit symmetries in the structure of the measurements and constraints. We believe our hierarchical regression and succinct operations to be of independent interest.

Within the Strongly Connected Components (SCCs) formed during the temporal evolution of a Cloud permission graph, we use the Burau Lyapunov exponent LE as an algebraic probe to locate the boundary between two risks regimes. We prove that no Abelian statistic (edge counts, net privilege flow, gate-firing rates) can determine LE. The non-commutation advantage is small, but actionable: we show how to leverage it to discriminate the two outstanding risk regimes, that we call dispersed and focused, for automating classification and governing remediation of risky Cloud permission flows.

OpenClaw-style agent stacks turn language into privileged execution: LLM intents flow through tool interception, policy gates, and a local executor. In parallel, skill marketplaces such as skills.sh make capability acquisition as easy as installing skills and CLIs, creating a growing capability supply chain. Together, these trends shift the dominant safety failure mode from "wrong answers" to execution-induced loss, where untrusted prompts, compromised skills, or narrative manipulation can trigger real trades and irreversible side effects. We propose Survivability-Aware Execution (SAE), an execution-layer survivability standard for OpenClaw-style systems and skill-enabled agents. SAE sits as middleware between a strategy engine (LLM or non-LLM) and the exchange executor. It defines an explicit execution contract (ExecutionRequest, ExecutionContext, ExecutionDecision) and enforces non-bypassable last-mile invariants: projection-based exposure budgets, cooldown and order-rate limits, slippage bounds, staged execution, and tool/venue allowlists. To make delegated execution testable under supply-chain risk, we operationalize the Delegation Gap (DG) via a logged Intended Policy Spec that enables deterministic out-of-scope labeling and reproducible DG metrics. On an offline replay using official Binance USD-M BTCUSDT/ETHUSDT perpetual data (15m; 2025-09-01–2025-12-01, incl. funding), SAE improves survivability: MDD drops from 0.4643 to 0.0319 (Full; 93.1%), |CVaR_0.99| shrinks from 4.025e-3 to ~1.02e-4 (~97.5%), and DG loss proxy falls from 0.647 to 0.019 (~97.0%). AttackSuccess decreases from 1.00 to 0.728 with zero FalseBlock in this run. Block bootstrap, paired Wilcoxon, and two-proportion tests confirm the shifts. SAE reframes agentic trading safety for the OpenClaw+skills era: treat upstream intent and skills as untrusted, and enforce survivability where actions become side effects.

The widespread adoption of thinking mode in large language models (LLMs) has significantly enhanced complex task processing capabilities while introducing new security risks. When subjected to jailbreak attacks, the step-by-step reasoning process may cause models to generate more detailed harmful content. We observe that thinking mode exhibits unique vulnerabilities when processing interleaved multiple tasks. Based on this observation, we propose multi-stream perturbation attack, which generates superimposed interference by interweaving multiple task streams within a single prompt. We design three perturbation strategies: multi-stream interleaving, inversion perturbation, and shape transformation, which disrupt the thinking process through concurrent task interleaving, character reversal, and format constraints respectively. On JailbreakBench, AdvBench, and HarmBench datasets, our method achieves attack success rates exceeding most methods across mainstream models including Qwen3 series, DeepSeek, Qwen3-Max, and Gemini 2.5 Flash. Experiments show thinking collapse rates and response repetition rates reach up to 17% and 60% respectively, indicating multi-stream perturbation not only bypasses safety mechanisms but also causes thinking process collapse or repetitive outputs.

Warning: This article includes red-teaming experiments, which contain examples of compromised LLM responses that may be offensive or upsetting. Large Language Models (LLMs) have the potential to create harmful content, such as generating sophisticated phishing emails and assisting in writing code of harmful computer viruses. Thus, it is crucial to ensure their safe and responsible response generation. To reduce the risk of generating harmful or irresponsible content, researchers have developed techniques such as reinforcement learning with human feedback to align LLM's outputs with human values and preferences. However, it is still undetermined whether such measures are sufficient to prevent LLMs from generating interesting responses. In this study, we propose Amnesia, a lightweight activation-space adversarial attack that manipulates internal transformer states to bypass existing safety mechanisms in open-weight LLMs. Through experimental analysis on state-of-the-art, open-weight LLMs, we demonstrate that our attack effectively circumvents existing safeguards, enabling the generation of harmful content without the need for any fine-tuning or additional training. Our experiments on benchmark datasets show that the proposed attack can induce various antisocial behaviors in LLMs. These findings highlight the urgent need for more robust security measures in open-weight LLMs and underscore the importance of continued research to prevent their potential misuse.

As backdoor attacks in UAV-based decentralized federated learning (DFL) grow increasingly stealthy and sophisticated, existing defenses have likewise escalated in complexity. Yet these defenses, which rely heavily on outlier detection, remain vulnerable to carefully crafted backdoors. In UAV-DFL, the lack of global coordination and limited resources further render outlier-based defenses impractical. Against this backdrop, gradient spectral analysis offers a promising alternative. While prior work primarily leverages low-frequency coefficients for pairwise comparisons, it neglects to analyze the intrinsic spectral characteristics of backdoor gradients. Through empirical analysis of existing stealthy attacks, we reveal a key insight: the more effort attackers invest in mimicking benign behaviors, the more distinct the spectral concentration becomes. Motivated by this, we propose Task-Aware Spectral Energy Refine (TASER) – a decentralized defense framework. To our knowledge, this is the first efficient backdoor defense that utilizes spectral concentration instead of complex outlier detection, enabling mitigation of stealthy attacks by structurally disrupting the backdoor task. To suppress the backdoor task, TASER preserves main-task-relevant frequency coefficients and discards others. We provide theoretical guarantees and demonstrate through experiments that TASER remains effective against stealthy backdoor attacks that bypass outlier-based defenses, achieving attack success rate below 20% and accuracy loss under 5%.

Large Language Models (LLMs) show promise for Automated Program Repair (APR), yet their effectiveness on security vulnerabilities remains poorly characterized. This study analyzes 319 LLM-generated security patchesacross 64 Java vulnerabilities from the Vul4J benchmark. Using tri-axis evaluation (compilation, security via PoV tests, functionality via test suites), the analysis reveals that only 24.8% of patches achieve full correctness, while 51.4% fail both security and functionality. The dominant failure mode is semantic misunderstanding: LLMs produce syntactically valid code but apply incorrect repair strategies. The proposed Security Repair Score (SRS) quantifies this gap, showing LLMs preserve functionality (mean 0.832) but struggle with security (mean 0.251). Vulnerability type strongly predicts difficulty, with fix rates ranging from 0% (input validation) to 45% (infinite loop). These findings demonstrate that LLM security patches require rigorous validation before deployment.

Most adversarial evaluations of large language model (LLM) safety assess single prompts and report binary pass/fail outcomes, which fails to capture how safety properties evolve under sustained adversarial interaction. We present ADVERSA, an automated red-teaming framework that measures guardrail degradation dynamics as continuous per-round compliance trajectories rather than discrete jailbreak events. ADVERSA uses a fine-tuned 70B attacker model (ADVERSA-Red, Llama-3.1-70B-Instruct with QLoRA) that eliminates the attacker-side safety refusals that render off-the-shelf models unreliable as attackers, scoring victim responses on a structured 5-point rubric that treats partial compliance as a distinct measurable state. We report a controlled experiment across three frontier victim models (Claude Opus 4.6, Gemini 3.1 Pro, GPT-5.2) using a triple-judge consensus architecture in which judge reliability is measured as a first-class research outcome rather than assumed. Across 15 conversations of up to 10 adversarial rounds, we observe a 26.7% jailbreak rate with an average jailbreak round of 1.25, suggesting that in this evaluation setting, successful jailbreaks were concentrated in early rounds rather than accumulating through sustained pressure. We document inter-judge agreement rates, self-judge scoring tendencies, attacker drift as a failure mode in fine-tuned attackers deployed out of their training distribution, and attacker refusals as a previously-underreported confound in victim resistance measurement. All limitations are stated explicitly. Attack prompts are withheld per responsible disclosure policy; all other experimental artifacts are released.

Role classification involves grouping hosts into related roles. It exposes the logical structure of a network, simplifies network management tasks such as policy checking and network segmentation, and can be used to improve the accuracy of network monitoring and analysis algorithms such as intrusion detection. This paper defines the role classification problem and introduces two practical algorithms that group hosts based on observed connection patterns while dealing with changes in these patterns over time. The algorithms have been implemented in a commercial network monitoring and analysis product for enterprise networks. Results from grouping two enterprise networks show that the number of groups identified by our algorithms can be two orders of magnitude smaller than the number of hosts and that the way our algorithms group hosts highly reflects the logical structure of the networks.

We explore the use of level structures to generalize the SQIsign signature scheme. We give a general framework where, given the public key and the commitment, the challenge is to exhibit an isogeny between them with an additional requirement, namely to map a chosen level structure to another. We then instantiate the framework using 1-dimensional and 2-dimensional isogenies. In doing that we provide a new explicit Deuring correspondence for supersingular elliptic curves with level structures and solve new constrained norm equations.

The temporal assumptions underpinning conventional Identity and Access Management collapse under agentic execution regimes. A sixty-second revocation window permits on the order of $6 \times 10^3$ unauthorized API calls at 100 ops/tick; at AWS Lambda scale, the figure approaches $6 \times 10^5$. This is a coherence problem, not merely a latency problem. We define a Capability Coherence System (CCS) and construct a state-mapping $\varphi : \Sigma_{\rm MESI} \to \Sigma_{\rm auth}$ preserving transition structure under bounded-staleness semantics. A safety theorem bounds unauthorized operations for the execution-count Release Consistency-directed Coherence (RCC) strategy at $D_{\rm rcc} \leq n$, independent of agent velocity $v$ – a qualitative departure from the $O(v \cdot \mathrm{TTL})$ scaling of time-bounded strategies. Tick-based discrete event simulation across three business-contextualised scenarios (four strategies, ten deterministic seeds each) confirms: RCC achieves a $120\times$ reduction versus TTL-based lease in the high-velocity scenario (50 vs. 6,000 unauthorized operations), and $184\times$ under anomaly-triggered revocation. Zero bound violations across all 120 runs confirm the per-capability safety guarantee. Simulation code: https://github.com/hipvlady/prizm

The widespread adoption of AI assistants has encouraged the development of privacy-aware platforms designed to extract insights from real-world usage. Their privacy protections primarily rely on layering multiple heuristic techniques, such as PII redaction, clustering, aggregation, and LLM-based privacy auditing. In this paper, we put their privacy claims to the test by presenting CLIOPATRA, the first attack against "privacy-preserving" LLM-based insights systems. Our attack involves an adversary that carefully designs and inserts malicious chats into the system to break multiple layers of protections and induce the leakage of sensitive information from a target user's chat. We evaluate CLIOPATRA on one such platform, Anthropic's Clio (now called "Anthropic Insights"), and target synthetically generated medical chats to show that an adversary can successfully and confidently (with nearly 100% precision) extract the medical history contained in these chats in up to 65% of cases. We also show that CLIOPATRA can stealthily extract information by obfuscating the private information in the generated insights. Finally, we demonstrate that existing ad hoc mitigations, such as LLM-based privacy auditing, are unreliable and fail to detect major leaks. Overall, our findings indicate that, even when layered, current heuristic protections are insufficient to adequately protect user data, and that prompt injection has been an understudied risk in LLM-based insight systems.

Current backdoor defenses assume that neutralizing a known trigger removes the backdoor. We show this trigger-centric view is incomplete: alternative triggers, patterns perceptually distinct from training triggers, reliably activate the same backdoor. We estimate the alternative trigger backdoor direction in feature space by contrasting clean and triggered representations, and then develop a feature-guided attack that jointly optimizes target prediction and directional alignment. First, we theoretically prove that alternative triggers exist and are an inevitable consequence of backdoor training. Then, we verify this empirically. Additionally, defenses that remove training triggers often leave backdoors intact, and alternative triggers can exploit the latent backdoor feature-space. Our findings motivate defenses targeting backdoor directions in representation space rather than input-space triggers.

Benchmarking presence-only passive reconnaissance in smart-grid communications is challenging because the adversary is receive-only, yet nearby observers can still alter propagation through additional shadowing and multipath that reshapes channel coherence. Public smart-grid cybersecurity datasets largely target active protocol- or measurement-layer attacks and rarely provide propagation-driven observables with tiered topology context, which limits reproducible evaluation under strictly passive threat models. This paper introduces an IEEE-inspired, literature-anchored benchmark dataset generator for passive reconnaissance over a tiered Home Area Network (HAN), Neighborhood Area Network (NAN), and Wide Area Network (WAN) communication graph with heterogeneous wireless and wireline links. Node-level time series are produced through a physically consistent channel-to-metrics mapping where channel state information (CSI) is represented via measurement-realistic amplitude and phase proxies that drive inferred signal-to-noise ratio (SNR), packet error behavior, and delay dynamics. Passive attacks are modeled only as windowed excess attenuation and coherence degradation with increased channel innovation, so reliability and latency deviations emerge through the same causal mapping without labels or feature shortcuts. The release provides split-independent realizations with burn-in removal, strictly causal temporal descriptors, adjacency-weighted neighbor aggregates and deviation features, and federated-ready per-node train, validation, and test partitions with train-only normalization metadata. Baseline federated experiments highlight technology-dependent detectability and enable standardized benchmarking of graph-temporal and federated detectors for passive reconnaissance.

The rapid expansion of Internet use has increased system exposure to cyber threats, with advanced persistent threats (APTs) being especially challenging due to their stealth, prolonged duration, and multi-stage attacks targeting high-value assets. In this study, we model APT evolution as a strategic interaction between an attacker and a defender on an attack graph. With limited information about the attacker's position and progress, the defender acts at random intervals by deploying intrusion detection sensors across the network. Once a compromise is detected, affected components are immediately secured through measures such as backdoor removal, patching, or system reconfiguration. Meanwhile, the attacker begins with reconnaissance and then proceeds through the network, exploiting vulnerabilities and installing backdoors to maintain persistent access and adaptive movement. Furthermore, the attacker may take several steps between consecutive defensive operations, resulting in an asymmetric temporal dynamic. The defender's goal is to reduce the likelihood that the attacker will gain access to a critical asset, whereas the attacker's purpose is to increase this likelihood. We investigate this interaction under three informational regimes, reflecting varying levels of attacker knowledge prior to action: (i) a Stackelberg scenario, in which the attacker has full knowledge of the defender's strategy and can optimize accordingly; (ii) a blind regime, where the attacker has no information and assumes uniform beliefs about defensive deployments; and (iii) a belief-based framework, where the attacker holds accurate probabilistic beliefs about the defender's actions. For each regime, we derive optimal defensive strategies by solving the corresponding optimization problems.

We establish the randomized distributed function computation (RDFC) framework, in which a sender transmits just enough information for a receiver to generate a randomized function of the input data. Describing RDFC as a form of semantic communication, which can be essentially seen as a generalized remote-source-coding problem, we show that security and privacy constraints naturally fit this model, as they generally require a randomization step. Using strong coordination metrics, we ensure (local differential) privacy for every input sequence and prove that such guarantees can be met even when no common randomness is shared between the transmitter and receiver. This work provides lower bounds on Wyner's common information (WCI), which is the communication cost when common randomness is absent, and proposes numerical techniques to evaluate the other corner point of the RDFC rate region for continuous-alphabet random variables with unlimited shared randomness. Experiments illustrate that a sufficient amount of common randomness can reduce the semantic communication rate by up to two orders of magnitude compared to the WCI point, while RDFC without any shared randomness still outperforms lossless transmission by a large margin. A finite blocklength analysis further confirms that the privacy parameter gap between the asymptotic and non-asymptotic RDFC methods closes exponentially fast with input length. Our results position RDFC as an energy-efficient semantic communication strategy for privacy-aware distributed computation systems.

Outsourcing encrypted data to the cloud creates a fundamental tension between data privacy and functional searchability. Current Searchable Symmetric Encryption (SSE) solutions frequently have significant limitations, such as excessive metadata leakage, or a lack of fine-grained access control. These issues restrict the scalability of secure searches in real-world applications where multiple clients require different levels of authorization. Our paper proposes MASSE, a dynamic multi-client SSE scheme incorporating attribute-based access control, which expands the OXT framework. With MASSE, clients are restricted sto searching for keywords authorized by their specific attribute sets, and the server remains unaware of the keywords and attributes. MASSE supports practical dynamic updates to documents, and client authorizations, including revocation, without requiring reencryption of the database or indices, or a large number of interactions. We formally prove the security of MASSE, that is, forward and backward privacy under a well-defined leakage profile, and token unforgeability. An experimental evaluation in a database containing 100 keywords, each associated with 150 documents, demonstrates the practical efficiency of MASSE. It takes less than two seconds to generate 10 to 100 keyword queries and 14 seconds to retrieve 50 matching documents. Theoretical results show that MASSE outperforms competing solutions, including OXT, and can be scaled to large encrypted databases. MASSE is also suitable for dynamic cloud deployments. Keywords: Searchable Encryption, SSE, Multi-Client, Attribute Based SSE, Access Control, Revocation, OXT

Software compartmentalization breaks down an application into compartments isolated from each other: an attacker taking over a compartment will be confined to it, limiting the damage they can cause to the rest of the application. Despite the security promises of this approach, recent studies have shown that most existing compartmentalized software is plagued by vulnerabilities at cross-compartment interfaces, allowing an attacker taking over a compartment to escape its confinement and negate the security guarantees expected from compartmentalization. In that context, securing cross-compartment interfaces is notoriously difficult and engineering-intensive. In light of recent advances in Automated Program Repair (APR), notably through the use of Large Language Models (LLMs), this paper presents a work in progress investigating the suitability of LLM-based APR at securing cross-compartment interfaces as automatically as possible. We observe that existing APR approaches and general purpose/code-centric LLMs used as is are unfit for this task, and present the design, implementation, and early results of a new APR framework dedicated to compartment interface safety. The framework integrates into a feedback loop 1) a specialized fuzzer uncovering cross-compartment interface vulnerabilities; 2) a patch generation component bridging the lack of compartmentalization awareness of existing LLMs with a series of analysis techniques; and 3) a patch validation component assessing the effectiveness of generated vulnerability fixes. We validate our framework over a sample interface vulnerability, comparing it to a naive use of general-purpose LLMs, and discuss future research avenues.

Diffusion models have made substantial advances in recent years, enabling high-quality image synthesis; however, the widespread dissemination and reuse of their outputs have introduced new challenges in intellectual property protection and content provenance. Image watermarking offers a solution to these challenges, and recent work has increasingly explored Noise-as-Watermark (NaW) approaches that integrate watermarking directly into the diffusion process. However, existing NaW methods fail to balance robustness and diversity. We attribute this weakness to value encoding, which encodes watermark bits into individual sampled values. It is extremely fragile in practical application scenarios. To address this, we encode watermark bits into the structured noise pattern, so that the watermark is preserved even when individual values are perturbed. To further ensure generation diversity, we introduce a dedicated randomization design that reshuffles the positions of noise elements without changing their values, preventing the watermark from inducing fixed noise patterns or spatial locations. Extensive experiments demonstrate that our method achieves state-of-the-art robustness while maintaining high generation quality across a wide range of lossy scenarios.

Analyzing Open Source Intelligence (OSINT) from large volumes of data is critical for drafting and publishing comprehensive CTI reports. This process usually follows a three-stage workflow – triage, deep search and TI drafting. While Large Language Models (LLMs) offer a promising route toward automation, existing benchmarks still have limitations. These benchmarks often consist of tasks that do not reflect real-world analyst workflows. For example, human analysts rarely receive tasks in the form of multiple-choice questions. Also, existing benchmarks often rely on model-centric metrics that emphasize lexical overlap rather than actionable, detailed insights essential for security analysts. Moreover, they typically fail to cover the complete three-stage workflow. To address these issues, we introduce CyberThreat-Eval, which is collected from the daily CTI workflow of a world-leading company. This expert-annotated benchmark assesses LLMs on practical tasks across all three stages as mentioned above. It utilizes analyst-centric metrics that measure factual accuracy, content quality, and operational costs. Our evaluation using this benchmark reveals important insights into the limitations of current LLMs. For example, LLMs often lack the nuanced expertise required to handle complex details and struggle to distinguish between correct and incorrect information. To address these challenges, the CTI workflow incorporates both external ground-truth databases and human expert knowledge. TRA allows human experts to iteratively provide feedback for continuous improvement. The code is available at \href{https://github.com/xschen-beb/CyberThreat-Eval}{GitHub} and \href{https://huggingface.co/datasets/xse/CyberThreat-Eval}{HuggingFace}.

The growth in the adoption of the WebAssembly (WASM) standard has given rise to a rapidly increasing landscape of binary applications that are natively ported to the environment of websites. The flexibility of WASM has made it the preferred way to run fast and resource-heavy applications, replacing a field that JavaScript previously monopolized. Despite its success, researchers have raised concerns over the security implementations of WASM, demonstrating that binary vulnerabilities, such as Buffer Overflows and Use After Free, remain a present danger for WASM binaries. Our work aims to demonstrate that such vulnerabilities, when occurring on a WebAssembly module, can affect the behavior of a web application in unexpected ways, enabling an attacker to exploit vulnerabilities that are typical of the web security landscape. We provide several scenarios to provide examples of how each binary vulnerability might lead to a web security vulnerability, such as SQL Injections, XS-Leaks, and SSTI. Our results show that binary vulnerabilities can invalidate common security mechanisms that web developer implement in their applications, demonstrating how the security of WASM modules remains a problem that needs to be addressed. We also provide a list of best practices and defensive strategies that developers can implement to mitigate the risks associated with running unsafe WASM modules in their web applications.

Tracking pixels are used to optimize online ad campaigns through personalization, re-targeting, and conversion tracking. Past research has primarily focused on detecting the prevalence of tracking pixels on the web, with limited attention to how they are configured across websites. A tracking pixel may be configured differently on different websites. In this paper, we present a differential analysis framework: PixelConfig, to reverse-engineer the configurations of Meta Pixel deployments across the web. Using this framework, we investigate three types of Meta Pixel configurations: activity tracking (i.e., what a user is doing on a website), identity tracking (i.e., who a user is or who the device is associated with), and tracking restrictions (i.e., mechanisms to limit the sharing of potentially sensitive information). Using data from the Internet Archive's Wayback Machine, we analyze and compare Meta Pixel configurations on 18K health-related websites with a control group of the top 10K websites from 2017 to 2024. We find that activity tracking features, such as automatic events that collect button clicks and page metadata, and identity tracking features, such as first-party cookies that are unaffected by third-party cookie blocking, reached adoption rates of up to 98.4%, largely driven by the Pixel's default settings. We also find that the Pixel is being used to track potentially sensitive information, such as user interactions related to booking medical appointments and button clicks associated with specific medical conditions (e.g., erectile dysfunction) on health-related websites. Tracking restriction features, such as Core Setup, are configured on up to 34.3% of health websites and 8.7% of control websites. However, even when enabled, these tracking restriction features provide limited protection and can be circumvented in practice.

Delegated quantum computation enables a client with limited quantum capabilities to outsource computations to a more powerful quantum server while preserving correctness and privacy. Verification is crucial in this setting to ensure that the untrusted quantum server performs the computation honestly and returns correct results. A common verification method is the quantum cut-and-choose technique. Inspired by classical verification methods for two-party computation, the client uses the majority of the delegated rounds to test the server's honesty, while keeping the remaining ones for the actual computation. Combining this technique with other methods, such as quantum error correction, could help achieve negligible cheating probabilities for the server; however, such methods can impose significant overheads making implementations unfeasible for the near-term future. In this work, we investigate whether cut-and-choose can yield efficient and secure verifiable quantum computation without additional costly techniques. We find that verifiable delegated quantum computation protocols relying solely on cut-and-choose techniques cannot be secure and efficient at the same time.

Advanced Persistent Threats (APTs) pose critical challenges to modern cybersecurity due to their multi-stage and stealthy nature. While provenance-based detection approaches show promise in capturing causal attack semantics, current threat provenance practices face two paradoxical issues: (1) expert skepticism, where human analysts doubt the capability of traditional detection models to identify complex attacks; and (2) expert dependence, as analysts cannot manually process large-scale raw logs to detect threats without these models. Consequently, collaboration between humans and traditional models remains the prevailing paradigm. However, this renders investigation quality contingent upon human expertise and frequently results in alert fatigue. To address these challenges, we present ProvAgent, a framework that evolves the threat provenance paradigm from human-model collaboration to a novel collaboration between multi-agent systems and traditional models. ProvAgent leverages the speed and cost-efficiency of traditional models for initial anomaly screening over large-scale logs. By enforcing fine-grained identity-behavior consistency via graph contrastive learning, it profiles entities based on specific attributes to generate high-fidelity alerts. With these alerts serving as investigation entry points, ProvAgent achieves in-depth autonomous investigation through a hypothesis-verification multi-agent framework. Evaluations with real-world datasets demonstrate that ProvAgent outperforms six state-of-the-art (SOTA) baselines in anomaly detection. Through automated investigation, ProvAgent reconstructs near-complete attack processes at a minimum cost of \$0.06 per day.

Dataset condensation (DC) learns a compact synthetic dataset that enables models to match the performance of full-data training, prioritising utility over distributional fidelity. While typically explored for computational efficiency, DC also holds promise for healthcare data democratisation, especially when paired with differential privacy, allowing synthetic data to serve as a safe alternative to real records. However, existing DC methods rely on differentiable neural networks, limiting their compatibility with widely used clinical models such as decision trees and Cox regression. We address this gap using a differentially private, zero-order optimisation framework that extends DC to non-differentiable models using only function evaluations. Empirical results across six datasets, including both classification and survival tasks, show that the proposed method produces condensed datasets that preserve model utility while providing effective differential privacy guarantees - enabling model-agnostic data sharing for clinical prediction tasks without exposing sensitive patient information.