{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:OGFLHOPSDG46IEFQLJNHBWKZZ5","short_pith_number":"pith:OGFLHOPS","schema_version":"1.0","canonical_sha256":"718ab3b9f219b9e410b05a5a70d959cf5925598f7a7d9ef66686b1ca899da9d0","source":{"kind":"arxiv","id":"2503.10945","version":3},"attestation_state":"computed","paper":{"title":"Gaussian DP for Reporting Differential Privacy Guarantees in Machine Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CR","stat.ML"],"primary_cat":"cs.LG","authors_text":"Antti Honkela, Bogdan Kulynych, Borja Balle, Flavio P. Calmon, Georgios Kaissis, Jamie Hayes, Juan Felipe Gomez","submitted_at":"2025-03-13T23:06:30Z","abstract_excerpt":"Current practices for reporting differential privacy (DP) guarantees for machine learning (ML) algorithms such as DP-SGD provide an incomplete and potentially misleading picture. For instance, if only a single $(\\varepsilon, \\delta)$ is known about a mechanism, standard analyses show that there could exist highly accurate inference attacks against training data records, when, upon a more careful analysis, such accurate attacks do not exist for most practical mechanisms. In this position paper, we argue that using _non-asymptotic_ Gaussian Differential Privacy (GDP) as the primary means of comm"},"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":"2503.10945","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-03-13T23:06:30Z","cross_cats_sorted":["cs.AI","cs.CR","stat.ML"],"title_canon_sha256":"291cd898809974dff8f5c9c12da221c549b37e50efe74225e41cc0778b72fc11","abstract_canon_sha256":"a80d497b1a275018f69c0e6ad9556ca04950e835011f907616079fd5e92ab0c6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-19T16:10:29.487638Z","signature_b64":"tN4mTgx3p5La+tkzXh3fup/TK2QV+Ws4RhXGDHV33OLO0xXOzDJ7Uu+7n6LFcoAVydQdsjCGWY4zIYSKkN+bDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"718ab3b9f219b9e410b05a5a70d959cf5925598f7a7d9ef66686b1ca899da9d0","last_reissued_at":"2026-06-19T16:10:29.487195Z","signature_status":"signed_v1","first_computed_at":"2026-06-19T16:10:29.487195Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Gaussian DP for Reporting Differential Privacy Guarantees in Machine Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CR","stat.ML"],"primary_cat":"cs.LG","authors_text":"Antti Honkela, Bogdan Kulynych, Borja Balle, Flavio P. Calmon, Georgios Kaissis, Jamie Hayes, Juan Felipe Gomez","submitted_at":"2025-03-13T23:06:30Z","abstract_excerpt":"Current practices for reporting differential privacy (DP) guarantees for machine learning (ML) algorithms such as DP-SGD provide an incomplete and potentially misleading picture. For instance, if only a single $(\\varepsilon, \\delta)$ is known about a mechanism, standard analyses show that there could exist highly accurate inference attacks against training data records, when, upon a more careful analysis, such accurate attacks do not exist for most practical mechanisms. In this position paper, we argue that using _non-asymptotic_ Gaussian Differential Privacy (GDP) as the primary means of comm"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.10945","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/2503.10945/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":"2503.10945","created_at":"2026-06-19T16:10:29.487252+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.10945v3","created_at":"2026-06-19T16:10:29.487252+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.10945","created_at":"2026-06-19T16:10:29.487252+00:00"},{"alias_kind":"pith_short_12","alias_value":"OGFLHOPSDG46","created_at":"2026-06-19T16:10:29.487252+00:00"},{"alias_kind":"pith_short_16","alias_value":"OGFLHOPSDG46IEFQ","created_at":"2026-06-19T16:10:29.487252+00:00"},{"alias_kind":"pith_short_8","alias_value":"OGFLHOPS","created_at":"2026-06-19T16:10:29.487252+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2602.08299","citing_title":"Cyclic Adaptive Private Synthesis for Sharing Real-World Data in Education","ref_index":22,"is_internal_anchor":true},{"citing_arxiv_id":"2605.14591","citing_title":"Privacy Auditing with Zero (0) Training Run","ref_index":17,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OGFLHOPSDG46IEFQLJNHBWKZZ5","json":"https://pith.science/pith/OGFLHOPSDG46IEFQLJNHBWKZZ5.json","graph_json":"https://pith.science/api/pith-number/OGFLHOPSDG46IEFQLJNHBWKZZ5/graph.json","events_json":"https://pith.science/api/pith-number/OGFLHOPSDG46IEFQLJNHBWKZZ5/events.json","paper":"https://pith.science/paper/OGFLHOPS"},"agent_actions":{"view_html":"https://pith.science/pith/OGFLHOPSDG46IEFQLJNHBWKZZ5","download_json":"https://pith.science/pith/OGFLHOPSDG46IEFQLJNHBWKZZ5.json","view_paper":"https://pith.science/paper/OGFLHOPS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.10945&json=true","fetch_graph":"https://pith.science/api/pith-number/OGFLHOPSDG46IEFQLJNHBWKZZ5/graph.json","fetch_events":"https://pith.science/api/pith-number/OGFLHOPSDG46IEFQLJNHBWKZZ5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OGFLHOPSDG46IEFQLJNHBWKZZ5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OGFLHOPSDG46IEFQLJNHBWKZZ5/action/storage_attestation","attest_author":"https://pith.science/pith/OGFLHOPSDG46IEFQLJNHBWKZZ5/action/author_attestation","sign_citation":"https://pith.science/pith/OGFLHOPSDG46IEFQLJNHBWKZZ5/action/citation_signature","submit_replication":"https://pith.science/pith/OGFLHOPSDG46IEFQLJNHBWKZZ5/action/replication_record"}},"created_at":"2026-06-19T16:10:29.487252+00:00","updated_at":"2026-06-19T16:10:29.487252+00:00"}