{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:IDANFVTJ72HIA73WYQZM3L4SP7","short_pith_number":"pith:IDANFVTJ","schema_version":"1.0","canonical_sha256":"40c0d2d669fe8e807f76c432cdaf927fca58658880b0dcd89ee5b9a27260ff5c","source":{"kind":"arxiv","id":"2112.03570","version":2},"attestation_state":"computed","paper":{"title":"Membership Inference Attacks From First Principles","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CR","authors_text":"Andreas Terzis, Florian Tramer, Milad Nasr, Nicholas Carlini, Shuang Song, Steve Chien","submitted_at":"2021-12-07T08:47:00Z","abstract_excerpt":"A membership inference attack allows an adversary to query a trained machine learning model to predict whether or not a particular example was contained in the model's training dataset. These attacks are currently evaluated using average-case \"accuracy\" metrics that fail to characterize whether the attack can confidently identify any members of the training set. We argue that attacks should instead be evaluated by computing their true-positive rate at low (e.g., <0.1%) false-positive rates, and find most prior attacks perform poorly when evaluated in this way. To address this we develop a Like"},"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":"2112.03570","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2021-12-07T08:47:00Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"9a464b12bef4ec50b327d464cefbdb4e848d969f154cd2ecc26635f79faa01ba","abstract_canon_sha256":"1d4418d6ff43885618b5c163dc710d0c6195c6b833ce32d30b60b0f3d8a5af06"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:13:47.216494Z","signature_b64":"5Mnn31PkFSq7lbVYbnO74Ik8CJ3lSpAMekOszO5V/BWxM9aCVsm0YX3Kb7maT+sT5f7Ys2AYKir6jzNPqhAXCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"40c0d2d669fe8e807f76c432cdaf927fca58658880b0dcd89ee5b9a27260ff5c","last_reissued_at":"2026-07-05T04:13:47.215929Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:13:47.215929Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Membership Inference Attacks From First Principles","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CR","authors_text":"Andreas Terzis, Florian Tramer, Milad Nasr, Nicholas Carlini, Shuang Song, Steve Chien","submitted_at":"2021-12-07T08:47:00Z","abstract_excerpt":"A membership inference attack allows an adversary to query a trained machine learning model to predict whether or not a particular example was contained in the model's training dataset. These attacks are currently evaluated using average-case \"accuracy\" metrics that fail to characterize whether the attack can confidently identify any members of the training set. We argue that attacks should instead be evaluated by computing their true-positive rate at low (e.g., <0.1%) false-positive rates, and find most prior attacks perform poorly when evaluated in this way. To address this we develop a Like"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2112.03570","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/2112.03570/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":"2112.03570","created_at":"2026-07-05T04:13:47.215996+00:00"},{"alias_kind":"arxiv_version","alias_value":"2112.03570v2","created_at":"2026-07-05T04:13:47.215996+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2112.03570","created_at":"2026-07-05T04:13:47.215996+00:00"},{"alias_kind":"pith_short_12","alias_value":"IDANFVTJ72HI","created_at":"2026-07-05T04:13:47.215996+00:00"},{"alias_kind":"pith_short_16","alias_value":"IDANFVTJ72HIA73W","created_at":"2026-07-05T04:13:47.215996+00:00"},{"alias_kind":"pith_short_8","alias_value":"IDANFVTJ","created_at":"2026-07-05T04:13:47.215996+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.05898","citing_title":"Auditing of Unlearning Algorithms","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2605.27825","citing_title":"MRMMIA: Membership Inference Attacks on Memory in Chat Agents","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30462","citing_title":"idSCD: Identifying Training Datasets through Semantic Correlation Descriptors","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2606.23591","citing_title":"Quantifying the Agreement Between Data-Influence and Data-Similarity to Understand LLM Behavior","ref_index":63,"is_internal_anchor":false},{"citing_arxiv_id":"2505.12546","citing_title":"Extracting memorized pieces of (copyrighted) books from open-weight language models","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21860","citing_title":"Robust Statistical Estimators with Bounded Empirical Sensitivity","ref_index":282,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IDANFVTJ72HIA73WYQZM3L4SP7","json":"https://pith.science/pith/IDANFVTJ72HIA73WYQZM3L4SP7.json","graph_json":"https://pith.science/api/pith-number/IDANFVTJ72HIA73WYQZM3L4SP7/graph.json","events_json":"https://pith.science/api/pith-number/IDANFVTJ72HIA73WYQZM3L4SP7/events.json","paper":"https://pith.science/paper/IDANFVTJ"},"agent_actions":{"view_html":"https://pith.science/pith/IDANFVTJ72HIA73WYQZM3L4SP7","download_json":"https://pith.science/pith/IDANFVTJ72HIA73WYQZM3L4SP7.json","view_paper":"https://pith.science/paper/IDANFVTJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2112.03570&json=true","fetch_graph":"https://pith.science/api/pith-number/IDANFVTJ72HIA73WYQZM3L4SP7/graph.json","fetch_events":"https://pith.science/api/pith-number/IDANFVTJ72HIA73WYQZM3L4SP7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IDANFVTJ72HIA73WYQZM3L4SP7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IDANFVTJ72HIA73WYQZM3L4SP7/action/storage_attestation","attest_author":"https://pith.science/pith/IDANFVTJ72HIA73WYQZM3L4SP7/action/author_attestation","sign_citation":"https://pith.science/pith/IDANFVTJ72HIA73WYQZM3L4SP7/action/citation_signature","submit_replication":"https://pith.science/pith/IDANFVTJ72HIA73WYQZM3L4SP7/action/replication_record"}},"created_at":"2026-07-05T04:13:47.215996+00:00","updated_at":"2026-07-05T04:13:47.215996+00:00"}