{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:OFSQXFBB72NIUFSDC4AQZIKZYT","short_pith_number":"pith:OFSQXFBB","schema_version":"1.0","canonical_sha256":"71650b9421fe9a8a164317010ca159c4f8e0f44d01be1afbd3f464f971e17801","source":{"kind":"arxiv","id":"2302.07956","version":1},"attestation_state":"computed","paper":{"title":"Tight Auditing of Differentially Private Machine Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.LG","authors_text":"Andreas Terzis, Borja Balle, Florian Tram\\`er, Jamie Hayes, Matthew Jagielski, Milad Nasr, Nicholas Carlini, Thomas Steinke","submitted_at":"2023-02-15T21:40:33Z","abstract_excerpt":"Auditing mechanisms for differential privacy use probabilistic means to empirically estimate the privacy level of an algorithm. For private machine learning, existing auditing mechanisms are tight: the empirical privacy estimate (nearly) matches the algorithm's provable privacy guarantee. But these auditing techniques suffer from two limitations. First, they only give tight estimates under implausible worst-case assumptions (e.g., a fully adversarial dataset). Second, they require thousands or millions of training runs to produce non-trivial statistical estimates of the privacy leakage.\n  This"},"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":"2302.07956","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-02-15T21:40:33Z","cross_cats_sorted":["cs.CR"],"title_canon_sha256":"b2c0b59aedc96ea09696da0053648ce8f2682c46bc291211278b9388aa303ef1","abstract_canon_sha256":"fd6f0811ca2d3dfc21b2add55484d8b8659b39a81f38ec1bb809f0af5340e0f9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:42:32.425950Z","signature_b64":"+8P1kZ1HtWQ1xwWJdfPNpMMXFB8Lwcqi/6UigiQffcvNlomW63YvQk3NUkkSwjH33KqSBLF/vpY6L01hyYEpCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"71650b9421fe9a8a164317010ca159c4f8e0f44d01be1afbd3f464f971e17801","last_reissued_at":"2026-07-05T05:42:32.425465Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:42:32.425465Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Tight Auditing of Differentially Private Machine Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.LG","authors_text":"Andreas Terzis, Borja Balle, Florian Tram\\`er, Jamie Hayes, Matthew Jagielski, Milad Nasr, Nicholas Carlini, Thomas Steinke","submitted_at":"2023-02-15T21:40:33Z","abstract_excerpt":"Auditing mechanisms for differential privacy use probabilistic means to empirically estimate the privacy level of an algorithm. For private machine learning, existing auditing mechanisms are tight: the empirical privacy estimate (nearly) matches the algorithm's provable privacy guarantee. But these auditing techniques suffer from two limitations. First, they only give tight estimates under implausible worst-case assumptions (e.g., a fully adversarial dataset). Second, they require thousands or millions of training runs to produce non-trivial statistical estimates of the privacy leakage.\n  This"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2302.07956","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/2302.07956/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":"2302.07956","created_at":"2026-07-05T05:42:32.425521+00:00"},{"alias_kind":"arxiv_version","alias_value":"2302.07956v1","created_at":"2026-07-05T05:42:32.425521+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2302.07956","created_at":"2026-07-05T05:42:32.425521+00:00"},{"alias_kind":"pith_short_12","alias_value":"OFSQXFBB72NI","created_at":"2026-07-05T05:42:32.425521+00:00"},{"alias_kind":"pith_short_16","alias_value":"OFSQXFBB72NIUFSD","created_at":"2026-07-05T05:42:32.425521+00:00"},{"alias_kind":"pith_short_8","alias_value":"OFSQXFBB","created_at":"2026-07-05T05:42:32.425521+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.27292","citing_title":"Detectability in Diversity: Improved Canary Crafting for Privacy Auditing in One Run","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2310.16789","citing_title":"Detecting Pretraining Data from Large Language Models","ref_index":107,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07674","citing_title":"Differentially Private Auditing Under Strategic Response","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OFSQXFBB72NIUFSDC4AQZIKZYT","json":"https://pith.science/pith/OFSQXFBB72NIUFSDC4AQZIKZYT.json","graph_json":"https://pith.science/api/pith-number/OFSQXFBB72NIUFSDC4AQZIKZYT/graph.json","events_json":"https://pith.science/api/pith-number/OFSQXFBB72NIUFSDC4AQZIKZYT/events.json","paper":"https://pith.science/paper/OFSQXFBB"},"agent_actions":{"view_html":"https://pith.science/pith/OFSQXFBB72NIUFSDC4AQZIKZYT","download_json":"https://pith.science/pith/OFSQXFBB72NIUFSDC4AQZIKZYT.json","view_paper":"https://pith.science/paper/OFSQXFBB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2302.07956&json=true","fetch_graph":"https://pith.science/api/pith-number/OFSQXFBB72NIUFSDC4AQZIKZYT/graph.json","fetch_events":"https://pith.science/api/pith-number/OFSQXFBB72NIUFSDC4AQZIKZYT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OFSQXFBB72NIUFSDC4AQZIKZYT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OFSQXFBB72NIUFSDC4AQZIKZYT/action/storage_attestation","attest_author":"https://pith.science/pith/OFSQXFBB72NIUFSDC4AQZIKZYT/action/author_attestation","sign_citation":"https://pith.science/pith/OFSQXFBB72NIUFSDC4AQZIKZYT/action/citation_signature","submit_replication":"https://pith.science/pith/OFSQXFBB72NIUFSDC4AQZIKZYT/action/replication_record"}},"created_at":"2026-07-05T05:42:32.425521+00:00","updated_at":"2026-07-05T05:42:32.425521+00:00"}