{"paper":{"title":"Content-Aware Attack Detection in LLM Agent Tool-Call Traffic: An Empirical Study of Features, Architectures, and Evaluation Protocols","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"MCPShield shows that content embeddings of tool arguments and responses are essential for detecting attacks on LLM agent traffic, lifting AUROC from 0.64 to above 0.89.","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CR","authors_text":"Sultan Zavrak","submitted_at":"2026-05-11T14:55:48Z","abstract_excerpt":"The Model Context Protocol (MCP) has become a widely adopted interface for LLM agents to invoke external tools, yet learned monitoring of MCP tool-call traffic remains underexplored. In this article, the proposed detector is presented as an attack detection framework for MCP tool-call traffic that encodes each agent session as a graph (tool calls as nodes, sequential and data-flow links as edges), enriches nodes with sentence-embedding features over arguments and responses, and classifies sessions as benign or attacked. Three GNN architectures (GAT, GCN, GraphSAGE), a no-graph MLP, and classic"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Content-level features are essential: metadata-only detection plateaus around an AUROC of 0.64 regardless of architecture, while content embeddings push the AUROC above 0.89.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The attack examples in RAS-Eval and ATBench are representative of real-world threats and that SBERT embeddings capture generalizable attack signals rather than dataset-specific artifacts.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"MCPShield detects attacks on LLM agent tool-call traffic by encoding sessions as graphs enriched with SBERT content embeddings, achieving AUROC above 0.89 with content features versus 0.64 for metadata alone.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"MCPShield shows that content embeddings of tool arguments and responses are essential for detecting attacks on LLM agent traffic, lifting AUROC from 0.64 to above 0.89.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"2be7d0554a6898f025b0c21336f8fcb86f458ff6dc2839868e23af16eb035329"},"source":{"id":"2605.11053","kind":"arxiv","version":3},"verdict":{"id":"18baac78-0ee7-4c11-9d03-a76e77dfd00c","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T21:13:06.877944Z","strongest_claim":"Content-level features are essential: metadata-only detection plateaus around an AUROC of 0.64 regardless of architecture, while content embeddings push the AUROC above 0.89.","one_line_summary":"MCPShield detects attacks on LLM agent tool-call traffic by encoding sessions as graphs enriched with SBERT content embeddings, achieving AUROC above 0.89 with content features versus 0.64 for metadata alone.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The attack examples in RAS-Eval and ATBench are representative of real-world threats and that SBERT embeddings capture generalizable attack signals rather than dataset-specific artifacts.","pith_extraction_headline":"MCPShield shows that content embeddings of tool arguments and responses are essential for detecting attacks on LLM agent traffic, lifting AUROC from 0.64 to above 0.89."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.11053/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"claim_evidence","ran_at":"2026-05-20T05:22:00.477103Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T14:38:23.932674Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T11:01:17.183645Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T09:04:11.242458Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"4029f4fd7ed3d4315e94a5eedda203f9c01f678d2ad697d8e7bca38106607601"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":1,"snapshot_sha256":"935fcd76a398a5376ac7bcde7de16c8ebef2edfa366fa86a3a77fd3f313afb2a"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}