268 papers
This paper proves that the conjectured conversion rule, which maps a Rényi Differential Privacy profile to an -Differential Privacy trade-off function via the pointwise maximum of single-order bounds (equivalent to the intersection of RDP privacy regions), is optimal and cannot be uniformly improved upon for any valid RDP profile or Type I error level.
This paper introduces Adaptive Tamaraw, a unified framework that clusters traffic traces to dynamically apply lightweight, set-specific padding parameters, thereby achieving provable security against website fingerprinting attacks while significantly reducing communication overhead compared to traditional defenses.
This paper introduces Kite, a protocol that enables private delegation of voting power in Decentralized Autonomous Organizations (DAOs) by allowing voters to delegate, revoke, or re-delegate their votes without revealing delegate identities to anyone, including the delegates themselves, while maintaining public verifiability and demonstrating practical implementation on Ethereum.
This paper introduces VisPoison, a backdoor attack framework that exploits text-to-visualization models for tabular data by using stealthy triggers to cause data exposure, misleading visualizations, or denial-of-service failures with over 90% success rates, thereby highlighting critical security vulnerabilities in current systems and the inadequacy of existing defenses.
This paper presents a longitudinal study of GPT, Llama, and Qwen models, revealing that continuous updates and increased model sizes do not consistently enhance adversarial robustness against misclassification, jailbreaks, and hallucinations, and can sometimes exacerbate existing vulnerabilities.
By combining data from a representative sample of 5,000 American adults with breach records from Have I Been Pwned, the study estimates that at least 82.84% of Americans have experienced an account breach, with an average of at least three breaches per person, a risk that is higher among women, White individuals, the middle-aged, and those with higher education levels.
This paper addresses the problem of role classification in enterprise networks by introducing two practical algorithms that group hosts based on evolving connection patterns to simplify network management and enhance monitoring accuracy, demonstrating their effectiveness through commercial implementation and significant reduction in host grouping complexity.
This paper proposes a Capability Coherence System (CCS) that maps memory consistency models to identity management, demonstrating through simulation that a Release Consistency-directed revocation strategy (RCC) achieves a constant bound on unauthorized operations independent of agent velocity, thereby outperforming traditional time-bounded approaches by orders of magnitude in high-speed agentic environments.
The paper introduces CLIOPATRA, a novel attack demonstrating that current heuristic privacy protections in LLM insight systems like Anthropic's Clio are insufficient, as a realistic adversary can successfully extract sensitive user medical history by injecting malicious chats, even with minimal prior knowledge of the target.
This paper challenges the assumption that neutralizing known triggers eliminates backdoors by demonstrating that perceptually distinct "alternative triggers" can reliably activate latent backdoor directions in feature space, thereby advocating for defenses that target these underlying representation patterns rather than specific input triggers.
This paper proposes a game-theoretic framework for modeling stealthy Advanced Persistent Threat (APT) defense on attack graphs, where a defender strategically deploys sensors to secure critical assets against an MDP-based attacker operating under asymmetric temporal dynamics and varying levels of informational knowledge.
The paper introduces MASSE, a dynamic multi-client Searchable Symmetric Encryption (SSE) scheme that extends the OXT framework to enable fine-grained, attribute-based access control and efficient updates without re-encryption, while maintaining forward and backward privacy and demonstrating practical scalability.
This paper presents a specialized, LLM-based Automated Program Repair framework designed to automatically secure cross-compartment interfaces in compartmentalized software by integrating a fuzzer, compartment-aware analysis techniques, and a validation component to address the limitations of existing general-purpose repair tools.
This paper introduces the non-commutative Group Ring Learning with Errors () problem based on semi-direct products of cyclic groups and establishes its cryptographic security by providing polynomial-time quantum reductions from the worst-case Shortest Independent Vectors Problem () to both the search and decision versions of .
ShapeMark is a robust and diversity-preserving watermarking method for diffusion models that encodes watermark bits into structured noise patterns rather than individual values and employs a randomization design to prevent fixed patterns, thereby achieving state-of-the-art robustness against lossy scenarios while maintaining high generation quality.
This paper demonstrates that binary vulnerabilities in WebAssembly modules, such as buffer overflows and use-after-free errors, can compromise web application security by enabling attacks like SQL injections and XS-Leaks, while also offering best practices to mitigate these risks.
This paper introduces PixelConfig, a framework for reverse-engineering Meta Pixel configurations, which reveals that default settings drive widespread adoption of activity and identity tracking features capable of capturing sensitive health data, while existing tracking restriction mechanisms offer limited practical protection.
ProvAgent is a novel framework that enhances threat detection and investigation by integrating graph contrastive learning for high-fidelity alert generation with a multi-agent collaborative system to autonomously reconstruct complex APT attack processes, thereby overcoming the limitations of traditional human-model collaboration.
This paper proposes a robust and provably secure image steganography framework that utilizes latent-space iterative optimization to significantly enhance message extraction accuracy under various compression and processing scenarios while maintaining security guarantees.
This paper presents a honeypot-based study conducted in early 2025 that reveals human Tor users primarily discover onion sites via the Ahmia search engine and exhibit significantly higher engagement with child sexual abuse material (CSAM) compared to other illicit categories, with English-language versions attracting the most interaction.