{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:EUXIG42XVCKGSS4ULMZKKWYHEL","short_pith_number":"pith:EUXIG42X","schema_version":"1.0","canonical_sha256":"252e837357a894694b945b32a55b0722ed63ba798d61bf48da3c6eb1a66a3186","source":{"kind":"arxiv","id":"2406.07954","version":1},"attestation_state":"computed","paper":{"title":"Dataset and Lessons Learned from the 2024 SaTML LLM Capture-the-Flag Competition","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Ahmed Salem, Chenhao Li, Daniel Paleka, Dragos Albastroiu, Edoardo Debenedetti, Florian Tram\\`er, Giovanni Cherubin, Javier Rando, Lea Sch\\\"onherr, Mario Fritz, Niv Cohen, Reshmi Ghosh, Robin Schmid, Rui Wen, Sahar Abdelnabi, Santiago Zanella-Beguelin, Silaghi Fineas Florin, Stefan Kraft, Takahiro Miki, Victor Klemm, Yuval Lemberg","submitted_at":"2024-06-12T07:27:28Z","abstract_excerpt":"Large language model systems face important security risks from maliciously crafted messages that aim to overwrite the system's original instructions or leak private data. To study this problem, we organized a capture-the-flag competition at IEEE SaTML 2024, where the flag is a secret string in the LLM system prompt. The competition was organized in two phases. In the first phase, teams developed defenses to prevent the model from leaking the secret. During the second phase, teams were challenged to extract the secrets hidden for defenses proposed by the other teams. This report summarizes the"},"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":"2406.07954","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2024-06-12T07:27:28Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"fc78426ef1e7f8c80921acc5376928b3fe7dfe7e1fc0271b78aa0b5db93d3585","abstract_canon_sha256":"317005cb47414a4c9c42781ceada379db2bf3e67be8dbe22c8488c9c60967af3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:30:41.267983Z","signature_b64":"WNAnqoPBQhDLWkrurnBPXDYu+5Fz/cgHtd+xIHttKKV6jaFsvDHQuJU321XbpbiMaH7ErdTeyuPwWv34dmW1AA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"252e837357a894694b945b32a55b0722ed63ba798d61bf48da3c6eb1a66a3186","last_reissued_at":"2026-07-05T08:30:41.267516Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:30:41.267516Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Dataset and Lessons Learned from the 2024 SaTML LLM Capture-the-Flag Competition","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Ahmed Salem, Chenhao Li, Daniel Paleka, Dragos Albastroiu, Edoardo Debenedetti, Florian Tram\\`er, Giovanni Cherubin, Javier Rando, Lea Sch\\\"onherr, Mario Fritz, Niv Cohen, Reshmi Ghosh, Robin Schmid, Rui Wen, Sahar Abdelnabi, Santiago Zanella-Beguelin, Silaghi Fineas Florin, Stefan Kraft, Takahiro Miki, Victor Klemm, Yuval Lemberg","submitted_at":"2024-06-12T07:27:28Z","abstract_excerpt":"Large language model systems face important security risks from maliciously crafted messages that aim to overwrite the system's original instructions or leak private data. To study this problem, we organized a capture-the-flag competition at IEEE SaTML 2024, where the flag is a secret string in the LLM system prompt. The competition was organized in two phases. In the first phase, teams developed defenses to prevent the model from leaking the secret. During the second phase, teams were challenged to extract the secrets hidden for defenses proposed by the other teams. This report summarizes the"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.07954","kind":"arxiv","version":1},"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/2406.07954/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":"2406.07954","created_at":"2026-07-05T08:30:41.267573+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.07954v1","created_at":"2026-07-05T08:30:41.267573+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.07954","created_at":"2026-07-05T08:30:41.267573+00:00"},{"alias_kind":"pith_short_12","alias_value":"EUXIG42XVCKG","created_at":"2026-07-05T08:30:41.267573+00:00"},{"alias_kind":"pith_short_16","alias_value":"EUXIG42XVCKGSS4U","created_at":"2026-07-05T08:30:41.267573+00:00"},{"alias_kind":"pith_short_8","alias_value":"EUXIG42X","created_at":"2026-07-05T08:30:41.267573+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2510.09093","citing_title":"Exploiting Web Search Tools of AI Agents for Data Exfiltration","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2406.13352","citing_title":"AgentDojo: A Dynamic Environment to Evaluate Prompt Injection Attacks and Defenses for LLM Agents","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/EUXIG42XVCKGSS4ULMZKKWYHEL","json":"https://pith.science/pith/EUXIG42XVCKGSS4ULMZKKWYHEL.json","graph_json":"https://pith.science/api/pith-number/EUXIG42XVCKGSS4ULMZKKWYHEL/graph.json","events_json":"https://pith.science/api/pith-number/EUXIG42XVCKGSS4ULMZKKWYHEL/events.json","paper":"https://pith.science/paper/EUXIG42X"},"agent_actions":{"view_html":"https://pith.science/pith/EUXIG42XVCKGSS4ULMZKKWYHEL","download_json":"https://pith.science/pith/EUXIG42XVCKGSS4ULMZKKWYHEL.json","view_paper":"https://pith.science/paper/EUXIG42X","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.07954&json=true","fetch_graph":"https://pith.science/api/pith-number/EUXIG42XVCKGSS4ULMZKKWYHEL/graph.json","fetch_events":"https://pith.science/api/pith-number/EUXIG42XVCKGSS4ULMZKKWYHEL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EUXIG42XVCKGSS4ULMZKKWYHEL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EUXIG42XVCKGSS4ULMZKKWYHEL/action/storage_attestation","attest_author":"https://pith.science/pith/EUXIG42XVCKGSS4ULMZKKWYHEL/action/author_attestation","sign_citation":"https://pith.science/pith/EUXIG42XVCKGSS4ULMZKKWYHEL/action/citation_signature","submit_replication":"https://pith.science/pith/EUXIG42XVCKGSS4ULMZKKWYHEL/action/replication_record"}},"created_at":"2026-07-05T08:30:41.267573+00:00","updated_at":"2026-07-05T08:30:41.267573+00:00"}