DRIFT: Dynamic Rule-Based Defense with Injection Isolation for Securing LLM Agents

Researchers introduce DRIFT, a new security framework designed to protect AI agents from prompt injection attacks through dynamic rule enforcement and memory isolation. The system uses a three-component approach with a Secure Planner, Dynamic Validator, and Injection Isolator to maintain security while preserving functionality across diverse AI models.

• →DRIFT addresses critical security vulnerabilities in LLM agents that interact with external environments and tools.
• →The framework introduces dynamic security policy updates, overcoming limitations of static defense systems.
• →Three-component architecture includes secure planning, dynamic validation, and injection isolation capabilities.
• →Empirical validation shows strong security performance while maintaining high utility across multiple benchmarks.
• →The solution targets risks including economic loss, privacy leakage, and system compromise from malicious inputs.