Cryptography & Security
Research on AI security, adversarial attacks, privacy, and cryptographic methods.
cs.CR · 505 papersEngineering Robustness into Personal Agents with the AI Workflow Store
This paper introduces an AI Workflow Store to integrate rigorous software engineering into AI agents, creating robust, reusable workflows instead of brittle on-the-fly systems.
Private Information Retrieval With Arbitrary Privacy Requirements for Graph-Based Storage
This paper redefines Private Information Retrieval (PIR) privacy for graph-based storage, enabling flexible, arbitrary requirements.
Local Private Information Retrieval: A New Privacy Perspective for Graph-Based Replicated Systems
This paper introduces "local PIR" for graph-replicated systems, showing significant communication efficiency gains by redefining user privacy.
BEACON: A Multimodal Dataset for Learning Behavioral Fingerprints from Gameplay Data
BEACON is a large, multimodal dataset from competitive Valorant gameplay for continuous authentication and behavioral fingerprinting research.
From Controlled to the Wild: Evaluation of Pentesting Agents for the Real-World
This paper introduces a new evaluation protocol for AI pentesting agents, shifting from task completion to realistic vulnerability discovery.
Democratizing Measurement of Critical Mobile Infrastructure: Security and Privacy in an Increasingly Centralized Communication Ecosystem
This paper introduces open-source measurement platforms to independently analyze the security and privacy of complex mobile communication ecosystems, including cellular and OTT services.
Threat Modelling using Domain-Adapted Language Models: Empirical Evaluation and Insights
An evaluation of domain-adapted LLMs for STRIDE threat modeling reveals inconsistent performance and fundamental limitations, urging task-specific reasoning.
LLMs for Secure Hardware Design and Related Problems: Opportunities and Challenges
A review of LLMs in hardware design, covering their capabilities, introduced vulnerabilities, and essential security countermeasures.
Can You Keep a Secret? Involuntary Information Leakage in Language Model Writing
Frontier language models involuntarily leak secret information thematically in their writing, even when instructed not to, posing a privacy risk.
LITMUS: Benchmarking Behavioral Jailbreaks of LLM Agents in Real OS Environments
LITMUS benchmarks LLM agent behavioral jailbreaks in real OS environments, revealing critical safety gaps and a new "Execution Hallucination" phenomenon.
MATRA: Modeling the Attack Surface of Agentic AI Systems -- OpenClaw Case Study
MATRA is a threat modeling framework for agentic AI systems, assessing risks and quantifying how architectural controls reduce attack blast radius.
AutoSOUP: Safety-Oriented Unit Proof Generation for Component-level Memory-Safety Verification
AutoSOUP automates component-level memory-safety verification using Safety-Oriented Unit Proofs and a hybrid LLM-as-function-call architecture.
diffGHOST: Diffusion based Generative Hedged Oblivious Synthetic Trajectories
diffGHOST is a diffusion model that generates privacy-preserving synthetic mobility trajectories by mitigating memorization in a segmented latent space.
Re-Triggering Safeguards within LLMs for Jailbreak Detection
This paper introduces an embedding disruption method to re-trigger LLM safeguards, effectively detecting and defending against jailbreak attacks.
Generate "Normal", Edit Poisoned: Branding Injection via Hint Embedding in Image Editing
This paper reveals a new vulnerability where hidden branding in images can be re-rendered by generative AI models during editing, even without explicit prompts.
Guaranteed Jailbreaking Defense via Disrupt-and-Rectify Smoothing
DR-Smoothing offers a guaranteed defense against LLM jailbreaking attacks by disrupting and rectifying prompts, balancing safety and helpfulness.
Acceptance Cards:A Four-Diagnostic Standard for Safe Fine-Tuning Defense Claims
Acceptance Cards introduce a four-diagnostic standard to rigorously evaluate safe fine-tuning defenses, revealing flaws in existing methods like SafeLoRA.
SoK: A Systematic Bidirectional Literature Review of AI & DLT Convergence
This paper systematically reviews AI and DLT convergence, classifying contributions and identifying neglected research areas and critical challenges.
Usability as a Weapon: Attacking the Safety of LLM-Based Code Generation via Usability Requirements
This paper introduces UPAttack, demonstrating how usability requirements can force LLMs to generate insecure code, achieving up to 98.1% attack success.
When Are LLM Inferences Acceptable? User Reactions and Control Preferences for Inferred Personal Information
Users are more curious than distressed by LLM inferences about personal data, with acceptability depending on context and third-party use.
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.