{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:MLQXUKTX6R5JP2A7YN3J2YHEA3","short_pith_number":"pith:MLQXUKTX","schema_version":"1.0","canonical_sha256":"62e17a2a77f47a97e81fc3769d60e406ddd0579aee9f996030a1719f895152ea","source":{"kind":"arxiv","id":"2412.18295","version":2},"attestation_state":"computed","paper":{"title":"Pirates of the RAG: Adaptively Attacking LLMs to Leak Knowledge Bases","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Christian Di Maio, Cristian Cosci, Marco Maggini, Stefano Melacci, Valentina Poggioni","submitted_at":"2024-12-24T09:03:57Z","abstract_excerpt":"The growing ubiquity of Retrieval-Augmented Generation (RAG) systems in several real-world services triggers severe concerns about their security. A RAG system improves the generative capabilities of a Large Language Models (LLM) by a retrieval mechanism which operates on a private knowledge base, whose unintended exposure could lead to severe consequences, including breaches of private and sensitive information. This paper presents a black-box attack to force a RAG system to leak its private knowledge base which, differently from existing approaches, is adaptive and automatic. A relevance-bas"},"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":"2412.18295","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2024-12-24T09:03:57Z","cross_cats_sorted":[],"title_canon_sha256":"00a288863ab787c2dd4d8dbe1ccd6a7b6e44182d48f14b66175a0c838b8036d3","abstract_canon_sha256":"1e6d0fb2783b083f76bfc73bd55892fc00ed622eace14cbc3b7aca6e0486edf8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:55:15.637017Z","signature_b64":"AN+f1tcfxF+zmiHmWjpBqfCc2M5Uk7O9rz2aN9uCP3i4+uVYHQMYSAj9oiRPoIeZ1ex02g/w6ZwoZEdTkt4mAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"62e17a2a77f47a97e81fc3769d60e406ddd0579aee9f996030a1719f895152ea","last_reissued_at":"2026-07-05T09:55:15.636548Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:55:15.636548Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Pirates of the RAG: Adaptively Attacking LLMs to Leak Knowledge Bases","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Christian Di Maio, Cristian Cosci, Marco Maggini, Stefano Melacci, Valentina Poggioni","submitted_at":"2024-12-24T09:03:57Z","abstract_excerpt":"The growing ubiquity of Retrieval-Augmented Generation (RAG) systems in several real-world services triggers severe concerns about their security. A RAG system improves the generative capabilities of a Large Language Models (LLM) by a retrieval mechanism which operates on a private knowledge base, whose unintended exposure could lead to severe consequences, including breaches of private and sensitive information. This paper presents a black-box attack to force a RAG system to leak its private knowledge base which, differently from existing approaches, is adaptive and automatic. A relevance-bas"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.18295","kind":"arxiv","version":2},"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/2412.18295/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":"2412.18295","created_at":"2026-07-05T09:55:15.636607+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.18295v2","created_at":"2026-07-05T09:55:15.636607+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.18295","created_at":"2026-07-05T09:55:15.636607+00:00"},{"alias_kind":"pith_short_12","alias_value":"MLQXUKTX6R5J","created_at":"2026-07-05T09:55:15.636607+00:00"},{"alias_kind":"pith_short_16","alias_value":"MLQXUKTX6R5JP2A7","created_at":"2026-07-05T09:55:15.636607+00:00"},{"alias_kind":"pith_short_8","alias_value":"MLQXUKTX","created_at":"2026-07-05T09:55:15.636607+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.18762","citing_title":"ALDEN: Boosting Private Data Extraction from Retrieval-Augmented Generation Systems via Active Learning and Distribution Estimation","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2510.06719","citing_title":"Differentially Private Synthetic Text Generation for Retrieval-Augmented Generation (RAG)","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2604.09747","citing_title":"ADAM: A Systematic Data Extraction Attack on Agent Memory via Adaptive Querying","ref_index":4,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MLQXUKTX6R5JP2A7YN3J2YHEA3","json":"https://pith.science/pith/MLQXUKTX6R5JP2A7YN3J2YHEA3.json","graph_json":"https://pith.science/api/pith-number/MLQXUKTX6R5JP2A7YN3J2YHEA3/graph.json","events_json":"https://pith.science/api/pith-number/MLQXUKTX6R5JP2A7YN3J2YHEA3/events.json","paper":"https://pith.science/paper/MLQXUKTX"},"agent_actions":{"view_html":"https://pith.science/pith/MLQXUKTX6R5JP2A7YN3J2YHEA3","download_json":"https://pith.science/pith/MLQXUKTX6R5JP2A7YN3J2YHEA3.json","view_paper":"https://pith.science/paper/MLQXUKTX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.18295&json=true","fetch_graph":"https://pith.science/api/pith-number/MLQXUKTX6R5JP2A7YN3J2YHEA3/graph.json","fetch_events":"https://pith.science/api/pith-number/MLQXUKTX6R5JP2A7YN3J2YHEA3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MLQXUKTX6R5JP2A7YN3J2YHEA3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MLQXUKTX6R5JP2A7YN3J2YHEA3/action/storage_attestation","attest_author":"https://pith.science/pith/MLQXUKTX6R5JP2A7YN3J2YHEA3/action/author_attestation","sign_citation":"https://pith.science/pith/MLQXUKTX6R5JP2A7YN3J2YHEA3/action/citation_signature","submit_replication":"https://pith.science/pith/MLQXUKTX6R5JP2A7YN3J2YHEA3/action/replication_record"}},"created_at":"2026-07-05T09:55:15.636607+00:00","updated_at":"2026-07-05T09:55:15.636607+00:00"}