{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:G5GYWNODQ5HFBIMKYZY5PK7MYX","short_pith_number":"pith:G5GYWNOD","schema_version":"1.0","canonical_sha256":"374d8b35c3874e50a18ac671d7abecc5f0bb6d5d0b6e1442f015fb464f2666a1","source":{"kind":"arxiv","id":"2109.11429","version":3},"attestation_state":"computed","paper":{"title":"Robin Hood and Matthew Effects: Differential Privacy Has Disparate Impact on Synthetic Data","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CR","cs.CY"],"primary_cat":"cs.LG","authors_text":"Bristena Oprisanu, Emiliano De Cristofaro, Georgi Ganev","submitted_at":"2021-09-23T15:14:52Z","abstract_excerpt":"Generative models trained with Differential Privacy (DP) can be used to generate synthetic data while minimizing privacy risks. We analyze the impact of DP on these models vis-a-vis underrepresented classes/subgroups of data, specifically, studying: 1) the size of classes/subgroups in the synthetic data, and 2) the accuracy of classification tasks run on them. We also evaluate the effect of various levels of imbalance and privacy budgets. Our analysis uses three state-of-the-art DP models (PrivBayes, DP-WGAN, and PATE-GAN) and shows that DP yields opposite size distributions in the generated s"},"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":"2109.11429","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-09-23T15:14:52Z","cross_cats_sorted":["cs.AI","cs.CR","cs.CY"],"title_canon_sha256":"90b8aa93e66b11814b74d6569572afb6186d38255faf98b762d625c21b908638","abstract_canon_sha256":"cae1fce90b1fa425836e0b9046f1202451c8ef32574bb56f37bbe9f2fef41cfb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:34:47.286800Z","signature_b64":"Qj42N144LIcbAQFP2IZaA/FKaiiyOuw0zHMeS0iaQs9itXNzzfVYRVQ+3qTfEJm3nQwd6dvp/yRKV2dzp2arBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"374d8b35c3874e50a18ac671d7abecc5f0bb6d5d0b6e1442f015fb464f2666a1","last_reissued_at":"2026-07-05T04:34:47.286203Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:34:47.286203Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Robin Hood and Matthew Effects: Differential Privacy Has Disparate Impact on Synthetic Data","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CR","cs.CY"],"primary_cat":"cs.LG","authors_text":"Bristena Oprisanu, Emiliano De Cristofaro, Georgi Ganev","submitted_at":"2021-09-23T15:14:52Z","abstract_excerpt":"Generative models trained with Differential Privacy (DP) can be used to generate synthetic data while minimizing privacy risks. We analyze the impact of DP on these models vis-a-vis underrepresented classes/subgroups of data, specifically, studying: 1) the size of classes/subgroups in the synthetic data, and 2) the accuracy of classification tasks run on them. We also evaluate the effect of various levels of imbalance and privacy budgets. Our analysis uses three state-of-the-art DP models (PrivBayes, DP-WGAN, and PATE-GAN) and shows that DP yields opposite size distributions in the generated s"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2109.11429","kind":"arxiv","version":3},"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/2109.11429/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":"2109.11429","created_at":"2026-07-05T04:34:47.286269+00:00"},{"alias_kind":"arxiv_version","alias_value":"2109.11429v3","created_at":"2026-07-05T04:34:47.286269+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2109.11429","created_at":"2026-07-05T04:34:47.286269+00:00"},{"alias_kind":"pith_short_12","alias_value":"G5GYWNODQ5HF","created_at":"2026-07-05T04:34:47.286269+00:00"},{"alias_kind":"pith_short_16","alias_value":"G5GYWNODQ5HFBIMK","created_at":"2026-07-05T04:34:47.286269+00:00"},{"alias_kind":"pith_short_8","alias_value":"G5GYWNOD","created_at":"2026-07-05T04:34:47.286269+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.10443","citing_title":"Tradeoffs in Privacy, Welfare, and Fairness for Facility Location","ref_index":22,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/G5GYWNODQ5HFBIMKYZY5PK7MYX","json":"https://pith.science/pith/G5GYWNODQ5HFBIMKYZY5PK7MYX.json","graph_json":"https://pith.science/api/pith-number/G5GYWNODQ5HFBIMKYZY5PK7MYX/graph.json","events_json":"https://pith.science/api/pith-number/G5GYWNODQ5HFBIMKYZY5PK7MYX/events.json","paper":"https://pith.science/paper/G5GYWNOD"},"agent_actions":{"view_html":"https://pith.science/pith/G5GYWNODQ5HFBIMKYZY5PK7MYX","download_json":"https://pith.science/pith/G5GYWNODQ5HFBIMKYZY5PK7MYX.json","view_paper":"https://pith.science/paper/G5GYWNOD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2109.11429&json=true","fetch_graph":"https://pith.science/api/pith-number/G5GYWNODQ5HFBIMKYZY5PK7MYX/graph.json","fetch_events":"https://pith.science/api/pith-number/G5GYWNODQ5HFBIMKYZY5PK7MYX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/G5GYWNODQ5HFBIMKYZY5PK7MYX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/G5GYWNODQ5HFBIMKYZY5PK7MYX/action/storage_attestation","attest_author":"https://pith.science/pith/G5GYWNODQ5HFBIMKYZY5PK7MYX/action/author_attestation","sign_citation":"https://pith.science/pith/G5GYWNODQ5HFBIMKYZY5PK7MYX/action/citation_signature","submit_replication":"https://pith.science/pith/G5GYWNODQ5HFBIMKYZY5PK7MYX/action/replication_record"}},"created_at":"2026-07-05T04:34:47.286269+00:00","updated_at":"2026-07-05T04:34:47.286269+00:00"}