AI Security
Papers in AI Security
16 papers
- The Promptware Kill Chain: How Prompt Injections Gradually Evolved Into a Multi-Step Malware
Structured framework to model and address evolving malware threats in LLM-based systems, termed promptware.
AI SecurityViability: 5.0 - BackdoorAgent: A Unified Framework for Backdoor Attacks on LLM-based Agents
A framework for identifying and analyzing backdoor threats in LLM-based agents, crucial for cybersecurity in AI workflows.
AI SecurityViability: 7.0 - Introducing the Generative Application Firewall (GAF)
Develop a unified generative firewall to secure LLM applications through a centralized enforcement layer.
AI SecurityViability: 2.0 - AgentGuardian: Learning Access Control Policies to Govern AI Agent Behavior
AgentGuardian secures AI agents with context-aware access control, preventing misuse and errors in real-time.
AI SecurityViability: 7.0 - Diffusion-Driven Deceptive Patches: Adversarial Manipulation and Forensic Detection in Facial Identity Verification
Commercialize a pipeline for adversarial patch creation and detection in facial recognition systems.
AI SecurityViability: 6.0 - Agent Skills in the Wild: An Empirical Study of Security Vulnerabilities at Scale
Developing a toolkit for detecting vulnerabilities in AI agent frameworks to enhance security.
AI SecurityViability: 4.0 - HardSecBench: Benchmarking the Security Awareness of LLMs for Hardware Code Generation
Benchmark tool to ensure security compliance of LLM-generated hardware and firmware code.
AI SecurityViability: 7.0 - Eliciting Harmful Capabilities by Fine-Tuning On Safeguarded Outputs
Reinforce AI safety by hardening models against elicitation attacks that exploit safeguarded outputs.
AI SecurityViability: 4.0 - Query-Efficient Agentic Graph Extraction Attacks on GraphRAG Systems
Develop a security tool that protects GraphRAG systems from efficient graph extraction attacks.
AI SecurityViability: 5.0 - Gaming the Judge: Unfaithful Chain-of-Thought Can Undermine Agent Evaluation
A system that secures LLM-based judgements by verifying agent reasoning against observable evidence.
AI SecurityViability: 4.0 - Don't believe everything you read: Understanding and Measuring MCP Behavior under Misleading Tool Descriptions
Develop a security auditing tool for MCP-based AI agents detecting description-code inconsistencies.
AI SecurityViability: 6.0 - Beyond Visual Safety: Jailbreaking Multimodal Large Language Models for Harmful Image Generation via Semantic-Agnostic Inputs
A framework for testing and revealing visual safety vulnerabilities in multimodal language models.
AI SecurityViability: 3.0 - Know Thy Enemy: Securing LLMs Against Prompt Injection via Diverse Data Synthesis and Instruction-Level Chain-of-Thought Learning
A security enhancement for LLMs that defends against prompt injection attacks using diverse data synthesis and instruction-level chain-of-thought learning.
AI SecurityViability: 7.0 - Defense Against Indirect Prompt Injection via Tool Result Parsing
A robust defense solution against indirect prompt injection attacks for LLM agents in autonomous systems, enhancing security with efficient tool result parsing.
AI SecurityViability: 7.0 - AgenticSCR: An Autonomous Agentic Secure Code Review for Immature Vulnerabilities Detection
AgenticSCR automates secure code review to catch immature vulnerabilities more accurately than traditional tools.
AI SecurityViability: 8.0 - RvB: Automating AI System Hardening via Iterative Red-Blue Games
RvB framework automates AI security system hardening through iterative red-blue team interactions.
AI SecurityViability: 5.0