{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:7KZ4DMKD2QZJJ5BCTWRGNO3XHM","short_pith_number":"pith:7KZ4DMKD","schema_version":"1.0","canonical_sha256":"fab3c1b143d43294f4229da266bb773b07758cd15e30da9913c89ff6625a55ff","source":{"kind":"arxiv","id":"2603.18382","version":2},"attestation_state":"computed","paper":{"title":"From Weak Cues to Real Identities: Evaluating Inference-Driven De-Anonymization in LLM Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Gyuhak Kim, Jihyun Jeong, Myeongseob Ko, Ruoxi Jia, Sumiran Singh Thakur","submitted_at":"2026-03-19T00:59:26Z","abstract_excerpt":"Anonymization is often assumed to protect privacy once explicit identifiers are removed, because re-identification has historically required specialized expertise, tailored algorithms, and manual corroboration. We show that LLM-based agents weaken this barrier: by combining scattered, individually non-identifying cues with public evidence, they reconstruct real-world identities, sometimes even during benign tasks. We evaluate this risk across three settings -- classical linkage incidents, a controlled benchmark (\\emph{InferLink}) that varies fingerprint type, task framing, and attacker knowled"},"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":"2603.18382","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-03-19T00:59:26Z","cross_cats_sorted":[],"title_canon_sha256":"21df51a5051209772cd654ad3977fe566f40e86bebaaf93a44e09cee310297f5","abstract_canon_sha256":"fdc4d32027aa056d0a671573e237e4545a05343998a592d8ce727dc0a0b199db"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-01T02:03:37.646475Z","signature_b64":"rLEKN1blGKQtKSB1ba3fq/XjoMyi5xvzscB/wbYFDgbC5npWvHbtrlpLuBIlAcxSEMlQSg1Poc69UBSPRCAADg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fab3c1b143d43294f4229da266bb773b07758cd15e30da9913c89ff6625a55ff","last_reissued_at":"2026-06-01T02:03:37.643844Z","signature_status":"signed_v1","first_computed_at":"2026-06-01T02:03:37.643844Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"From Weak Cues to Real Identities: Evaluating Inference-Driven De-Anonymization in LLM Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Gyuhak Kim, Jihyun Jeong, Myeongseob Ko, Ruoxi Jia, Sumiran Singh Thakur","submitted_at":"2026-03-19T00:59:26Z","abstract_excerpt":"Anonymization is often assumed to protect privacy once explicit identifiers are removed, because re-identification has historically required specialized expertise, tailored algorithms, and manual corroboration. We show that LLM-based agents weaken this barrier: by combining scattered, individually non-identifying cues with public evidence, they reconstruct real-world identities, sometimes even during benign tasks. We evaluate this risk across three settings -- classical linkage incidents, a controlled benchmark (\\emph{InferLink}) that varies fingerprint type, task framing, and attacker knowled"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.18382","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/2603.18382/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":"2603.18382","created_at":"2026-06-01T02:03:37.645243+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.18382v2","created_at":"2026-06-01T02:03:37.645243+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.18382","created_at":"2026-06-01T02:03:37.645243+00:00"},{"alias_kind":"pith_short_12","alias_value":"7KZ4DMKD2QZJ","created_at":"2026-06-01T02:03:37.645243+00:00"},{"alias_kind":"pith_short_16","alias_value":"7KZ4DMKD2QZJJ5BC","created_at":"2026-06-01T02:03:37.645243+00:00"},{"alias_kind":"pith_short_8","alias_value":"7KZ4DMKD","created_at":"2026-06-01T02:03:37.645243+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.14113","citing_title":"ProtoMedAgent: Multimodal Clinical Interpretability via Privacy-Aware Agentic Workflows","ref_index":18,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7KZ4DMKD2QZJJ5BCTWRGNO3XHM","json":"https://pith.science/pith/7KZ4DMKD2QZJJ5BCTWRGNO3XHM.json","graph_json":"https://pith.science/api/pith-number/7KZ4DMKD2QZJJ5BCTWRGNO3XHM/graph.json","events_json":"https://pith.science/api/pith-number/7KZ4DMKD2QZJJ5BCTWRGNO3XHM/events.json","paper":"https://pith.science/paper/7KZ4DMKD"},"agent_actions":{"view_html":"https://pith.science/pith/7KZ4DMKD2QZJJ5BCTWRGNO3XHM","download_json":"https://pith.science/pith/7KZ4DMKD2QZJJ5BCTWRGNO3XHM.json","view_paper":"https://pith.science/paper/7KZ4DMKD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.18382&json=true","fetch_graph":"https://pith.science/api/pith-number/7KZ4DMKD2QZJJ5BCTWRGNO3XHM/graph.json","fetch_events":"https://pith.science/api/pith-number/7KZ4DMKD2QZJJ5BCTWRGNO3XHM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7KZ4DMKD2QZJJ5BCTWRGNO3XHM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7KZ4DMKD2QZJJ5BCTWRGNO3XHM/action/storage_attestation","attest_author":"https://pith.science/pith/7KZ4DMKD2QZJJ5BCTWRGNO3XHM/action/author_attestation","sign_citation":"https://pith.science/pith/7KZ4DMKD2QZJJ5BCTWRGNO3XHM/action/citation_signature","submit_replication":"https://pith.science/pith/7KZ4DMKD2QZJJ5BCTWRGNO3XHM/action/replication_record"}},"created_at":"2026-06-01T02:03:37.645243+00:00","updated_at":"2026-06-01T02:03:37.645243+00:00"}