Data Leakage Prevention in Agentic Applications via Preemptive Hardening
Akansha Shukla, Emily Bellov, Parth Atulbhai Gandhi, Yuval Elovici, Asaf Shabtai
cs.CR, cs.AI
Submitted: 2026-07-21
Code: https://github.com/langchain-ai/langchain
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation
- Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection
- Identifying the Risks of LM Agents with an LM-Emulated Sandbox
- A Survey on the Safety and Security Threats of Computer-Using Agents: JARVIS or Ultron?
- "Do Not Mention This to the User": Detecting and Understanding Malicious Agent Skills in the Wild
- Fault-Tolerant Sandboxing for AI Coding Agents: A Transactional Approach to Safe Autonomous Execution
- Proactive defense against LLM Jailbreak
- Defending Against Prompt Injection With a Few DefensiveTokens
- Defense Against Prompt Injection Attack by Leveraging Attack Techniques
- StruQ: Defending Against Prompt Injection with Structured Queries
- PromptShield: Deployable Detection for Prompt Injection Attacks
- LlamaFirewall: An open source guardrail system for building secure AI agents
- Securing AI Agents with Information-Flow Control
- Prompt Flow Integrity to Prevent Privilege Escalation in LLM Agents
- Permissive Information-Flow Analysis for Large Language Models
- RTBAS: Defending LLM Agents Against Prompt Injection and Privacy Leakage
- AGrail: A Lifelong Agent Guardrail with Effective and Adaptive Safety Detection
- Progent: Securing AI Agents with Privilege Control
- Defeating Prompt Injections by Design
- IsolateGPT: An Execution Isolation Architecture for LLM-Based Agentic Systems
Related papers
- SoK: AI-Augmented Binary Reversing
- Relaxed Sender Anonymity for CBDC Interbank Settlement: A Zero-Knowledge Approach on Permissioned EVM
- Calibration-Family Overfit: Why Trusted Sabotage Monitors Don't Transfer Across Lineages
- Efficient Fuzzy PSI under One-Sided Assumptions
- Sealing the Audit-Runtime Gap for LLM Skills
- Token Composition: A Graph Based on EVM Logs