Backend
ToFu: Exploiting and Fuzzing the LLM-Orchestrator-Tool Trust Boundary
arXiv cs.CR0 views
arXiv:2610.13651v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly deployed as agents that interact with external software through tools, with an orchestration layer translating model outputs into structured tool invocations. This architecture creates an LLM-orchestrator-tool trust boundary where tool semantics may go unenforced and LLM-generated arguments may be incorrectly treated as trusted. We present ToFu, a tool-aware framework that automatically derives semantic security checks for LLM tools from their existing metadata and a CWE-guided sanitizer corpus. ToFu specializes these checks to individual tools and automatically generates fuzzing harnesses and semantic oracles using runtime introspection and outcome sanitization. We evaluated ToFu across 273 MCP servers containing 3405 tools and identified verified bugs corresponding to CWE weakness classes in 343 tools across 118 of the servers. Our findings show that real-world tools frequently fail to enforce their documented input semantics, making it unsafe to treat LLM-generated arguments as trusted.
Source：https://arxiv.org/abs/2610.13651PreviousAPEX-Harm: Evaluating Agent Safety in Shared Professional WorkspacesNextDecision-Focused Learning for Cyber-Physical Risk-Aware Power Grid Operations
Related posts
find -size -1M Matched Only the Empty File: Size Units in a Camera Clip Cleanup Script
0dev.to
Mastering Background Tasks in Flutter: Efficient Data Sync & UI Updates
0dev.to
Remove a Photo Background in the Browser: An On-Device Model, No Upload
0dev.to
My morning terminal: zoxide jumps, mise runs the project, starship shows context, fzf picks the file
0dev.to
A stale translation passes every check and still tells a different story
0dev.to
NYT and publishers sue Brave over AI content scraping
0dev.to