{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:QQKG6T5YKT4HPLNBXCNW2UGCZB","short_pith_number":"pith:QQKG6T5Y","schema_version":"1.0","canonical_sha256":"84146f4fb854f877ada1b89b6d50c2c86a611c1c663ac3f76b31752d021f48e9","source":{"kind":"arxiv","id":"2410.17401","version":4},"attestation_state":"computed","paper":{"title":"AdvAgent: Controllable Blackbox Red-teaming on Web Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CR","authors_text":"Bo Li, Chejian Xu, Huan Sun, Jiawei Zhang, Lingbo Mo, Mengqi Yuan, Mintong Kang, Zeyi Liao","submitted_at":"2024-10-22T20:18:26Z","abstract_excerpt":"Foundation model-based agents are increasingly used to automate complex tasks, enhancing efficiency and productivity. However, their access to sensitive resources and autonomous decision-making also introduce significant security risks, where successful attacks could lead to severe consequences. To systematically uncover these vulnerabilities, we propose AdvAgent, a black-box red-teaming framework for attacking web agents. Unlike existing approaches, AdvAgent employs a reinforcement learning-based pipeline to train an adversarial prompter model that optimizes adversarial prompts using feedback"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2410.17401","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2024-10-22T20:18:26Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"9c9741cf37ba1c2f8a4e95e67e1a45bdfe71c2cede3c085eb982514404186f9b","abstract_canon_sha256":"db9da53eee56a787822a0a5ddc036f213f194de2977b3d47bc7ad61ab3c9b9d8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:13:14.525990Z","signature_b64":"d3sPsPhRyUxODsfVrc0H9pGKeKqMlXOVHJvm3Y+cmnOXXnx7FnSkwCVU2U78tqO6mOhcc+2mvGpocXrEclq/Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"84146f4fb854f877ada1b89b6d50c2c86a611c1c663ac3f76b31752d021f48e9","last_reissued_at":"2026-07-05T11:13:14.525447Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:13:14.525447Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"AdvAgent: Controllable Blackbox Red-teaming on Web Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CR","authors_text":"Bo Li, Chejian Xu, Huan Sun, Jiawei Zhang, Lingbo Mo, Mengqi Yuan, Mintong Kang, Zeyi Liao","submitted_at":"2024-10-22T20:18:26Z","abstract_excerpt":"Foundation model-based agents are increasingly used to automate complex tasks, enhancing efficiency and productivity. However, their access to sensitive resources and autonomous decision-making also introduce significant security risks, where successful attacks could lead to severe consequences. To systematically uncover these vulnerabilities, we propose AdvAgent, a black-box red-teaming framework for attacking web agents. Unlike existing approaches, AdvAgent employs a reinforcement learning-based pipeline to train an adversarial prompter model that optimizes adversarial prompts using feedback"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.17401","kind":"arxiv","version":4},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2410.17401/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2410.17401","created_at":"2026-07-05T11:13:14.525514+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.17401v4","created_at":"2026-07-05T11:13:14.525514+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.17401","created_at":"2026-07-05T11:13:14.525514+00:00"},{"alias_kind":"pith_short_12","alias_value":"QQKG6T5YKT4H","created_at":"2026-07-05T11:13:14.525514+00:00"},{"alias_kind":"pith_short_16","alias_value":"QQKG6T5YKT4HPLNB","created_at":"2026-07-05T11:13:14.525514+00:00"},{"alias_kind":"pith_short_8","alias_value":"QQKG6T5Y","created_at":"2026-07-05T11:13:14.525514+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":11,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.08147","citing_title":"Prismata: Confining Cross-Site Prompt Injection in Web Agents","ref_index":88,"is_internal_anchor":true},{"citing_arxiv_id":"2606.20717","citing_title":"MIRAGE: Stealthy Visual Prompt Injection for Vulnerability Detection in Web Agents","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02449","citing_title":"HLL: Can Agents Cross Humanity's Last Line of Verification?","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15030","citing_title":"WARD: Adversarially Robust Defense of Web Agents Against Prompt Injections","ref_index":72,"is_internal_anchor":false},{"citing_arxiv_id":"2504.11703","citing_title":"Progent: Securing AI Agents with Privilege Control","ref_index":70,"is_internal_anchor":false},{"citing_arxiv_id":"2505.10924","citing_title":"A Survey on the Safety and Security Threats of Computer-Using Agents: JARVIS or Ultron?","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2506.23978","citing_title":"LLM Agents Are the Antidote to Walled Gardens","ref_index":101,"is_internal_anchor":false},{"citing_arxiv_id":"2507.04227","citing_title":"Mobile GUI Agents under Real-world Threats: Are We There Yet?","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2510.23883","citing_title":"Agentic AI Security: Threats, Defenses, Evaluation, and Open Challenges","ref_index":70,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06811","citing_title":"SkillTrojan: Backdoor Attacks on Skill-Based Agent Systems","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04426","citing_title":"ShieldNet: Network-Level Guardrails against Emerging Supply-Chain Injections in Agentic Systems","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QQKG6T5YKT4HPLNBXCNW2UGCZB","json":"https://pith.science/pith/QQKG6T5YKT4HPLNBXCNW2UGCZB.json","graph_json":"https://pith.science/api/pith-number/QQKG6T5YKT4HPLNBXCNW2UGCZB/graph.json","events_json":"https://pith.science/api/pith-number/QQKG6T5YKT4HPLNBXCNW2UGCZB/events.json","paper":"https://pith.science/paper/QQKG6T5Y"},"agent_actions":{"view_html":"https://pith.science/pith/QQKG6T5YKT4HPLNBXCNW2UGCZB","download_json":"https://pith.science/pith/QQKG6T5YKT4HPLNBXCNW2UGCZB.json","view_paper":"https://pith.science/paper/QQKG6T5Y","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.17401&json=true","fetch_graph":"https://pith.science/api/pith-number/QQKG6T5YKT4HPLNBXCNW2UGCZB/graph.json","fetch_events":"https://pith.science/api/pith-number/QQKG6T5YKT4HPLNBXCNW2UGCZB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QQKG6T5YKT4HPLNBXCNW2UGCZB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QQKG6T5YKT4HPLNBXCNW2UGCZB/action/storage_attestation","attest_author":"https://pith.science/pith/QQKG6T5YKT4HPLNBXCNW2UGCZB/action/author_attestation","sign_citation":"https://pith.science/pith/QQKG6T5YKT4HPLNBXCNW2UGCZB/action/citation_signature","submit_replication":"https://pith.science/pith/QQKG6T5YKT4HPLNBXCNW2UGCZB/action/replication_record"}},"created_at":"2026-07-05T11:13:14.525514+00:00","updated_at":"2026-07-05T11:13:14.525514+00:00"}