{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2022:IBL4AV5YBSUUBMY7XCMVB3PF2U","short_pith_number":"pith:IBL4AV5Y","canonical_record":{"source":{"id":"2210.03843","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-10-07T22:59:00Z","cross_cats_sorted":["cs.CR"],"title_canon_sha256":"57c3602b96eec67862c80a89003e96508e7036d79af2a2247fa1c5135a144ea8","abstract_canon_sha256":"8f25b1d6eb13070574344783a5607551120da9005d40dbd1969eb266c0fca83d"},"schema_version":"1.0"},"canonical_sha256":"4057c057b80ca940b31fb89950ede5d5181d97ddf00ca799ac7ebe6b74b6ac74","source":{"kind":"arxiv","id":"2210.03843","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2210.03843","created_at":"2026-07-05T05:04:32Z"},{"alias_kind":"arxiv_version","alias_value":"2210.03843v1","created_at":"2026-07-05T05:04:32Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2210.03843","created_at":"2026-07-05T05:04:32Z"},{"alias_kind":"pith_short_12","alias_value":"IBL4AV5YBSUU","created_at":"2026-07-05T05:04:32Z"},{"alias_kind":"pith_short_16","alias_value":"IBL4AV5YBSUUBMY7","created_at":"2026-07-05T05:04:32Z"},{"alias_kind":"pith_short_8","alias_value":"IBL4AV5Y","created_at":"2026-07-05T05:04:32Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2022:IBL4AV5YBSUUBMY7XCMVB3PF2U","target":"record","payload":{"canonical_record":{"source":{"id":"2210.03843","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-10-07T22:59:00Z","cross_cats_sorted":["cs.CR"],"title_canon_sha256":"57c3602b96eec67862c80a89003e96508e7036d79af2a2247fa1c5135a144ea8","abstract_canon_sha256":"8f25b1d6eb13070574344783a5607551120da9005d40dbd1969eb266c0fca83d"},"schema_version":"1.0"},"canonical_sha256":"4057c057b80ca940b31fb89950ede5d5181d97ddf00ca799ac7ebe6b74b6ac74","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:04:32.922793Z","signature_b64":"z1xqrW5YcmPuIGoy3wiGycALOtAh4/pcNHEFZJG9d6iFACrXjtR03Iq5UCqjqmtU6ZDWi73VtkpeHiYlzBhFDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4057c057b80ca940b31fb89950ede5d5181d97ddf00ca799ac7ebe6b74b6ac74","last_reissued_at":"2026-07-05T05:04:32.922428Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:04:32.922428Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2210.03843","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T05:04:32Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"AgQZ8i3WoqU0Mahf8CyitdddYZ2zOUc63KVJwzfZ1MmSSnNELSpPFJc+ruPfJ10V5teALMIZxA1tS+J/OpgcDQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-16T06:07:56.549177Z"},"content_sha256":"dbd829218f5036d8d9e06242d8c597319eecf4b2cc1b4526e5571cb34bec8d83","schema_version":"1.0","event_id":"sha256:dbd829218f5036d8d9e06242d8c597319eecf4b2cc1b4526e5571cb34bec8d83"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2022:IBL4AV5YBSUUBMY7XCMVB3PF2U","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Differentially Private Deep Learning with ModelMix","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.LG","authors_text":"Hanshen Xiao, Jun Wan, Srinivas Devadas","submitted_at":"2022-10-07T22:59:00Z","abstract_excerpt":"Training large neural networks with meaningful/usable differential privacy security guarantees is a demanding challenge. In this paper, we tackle this problem by revisiting the two key operations in Differentially Private Stochastic Gradient Descent (DP-SGD): 1) iterative perturbation and 2) gradient clipping. We propose a generic optimization framework, called {\\em ModelMix}, which performs random aggregation of intermediate model states. It strengthens the composite privacy analysis utilizing the entropy of the training trajectory and improves the $(\\epsilon, \\delta)$ DP security parameters "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2210.03843","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/2210.03843/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T05:04:32Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"9Ag1db5FlxwOct4tymVkqfq8EeZ0raf2jV0T5OGWcXbiE1Eo53q/35e5NeyzfpnY0C1hsHqoHFCJZC4zewc5CA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-16T06:07:56.550052Z"},"content_sha256":"ca7d07e53bb988095cf6423859f3d5cf9b58eb88769a3f8489c716e92577769e","schema_version":"1.0","event_id":"sha256:ca7d07e53bb988095cf6423859f3d5cf9b58eb88769a3f8489c716e92577769e"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/IBL4AV5YBSUUBMY7XCMVB3PF2U/bundle.json","state_url":"https://pith.science/pith/IBL4AV5YBSUUBMY7XCMVB3PF2U/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/IBL4AV5YBSUUBMY7XCMVB3PF2U/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-16T06:07:56Z","links":{"resolver":"https://pith.science/pith/IBL4AV5YBSUUBMY7XCMVB3PF2U","bundle":"https://pith.science/pith/IBL4AV5YBSUUBMY7XCMVB3PF2U/bundle.json","state":"https://pith.science/pith/IBL4AV5YBSUUBMY7XCMVB3PF2U/state.json","well_known_bundle":"https://pith.science/.well-known/pith/IBL4AV5YBSUUBMY7XCMVB3PF2U/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2022:IBL4AV5YBSUUBMY7XCMVB3PF2U","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"8f25b1d6eb13070574344783a5607551120da9005d40dbd1969eb266c0fca83d","cross_cats_sorted":["cs.CR"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-10-07T22:59:00Z","title_canon_sha256":"57c3602b96eec67862c80a89003e96508e7036d79af2a2247fa1c5135a144ea8"},"schema_version":"1.0","source":{"id":"2210.03843","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2210.03843","created_at":"2026-07-05T05:04:32Z"},{"alias_kind":"arxiv_version","alias_value":"2210.03843v1","created_at":"2026-07-05T05:04:32Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2210.03843","created_at":"2026-07-05T05:04:32Z"},{"alias_kind":"pith_short_12","alias_value":"IBL4AV5YBSUU","created_at":"2026-07-05T05:04:32Z"},{"alias_kind":"pith_short_16","alias_value":"IBL4AV5YBSUUBMY7","created_at":"2026-07-05T05:04:32Z"},{"alias_kind":"pith_short_8","alias_value":"IBL4AV5Y","created_at":"2026-07-05T05:04:32Z"}],"graph_snapshots":[{"event_id":"sha256:ca7d07e53bb988095cf6423859f3d5cf9b58eb88769a3f8489c716e92577769e","target":"graph","created_at":"2026-07-05T05:04:32Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2210.03843/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Training large neural networks with meaningful/usable differential privacy security guarantees is a demanding challenge. In this paper, we tackle this problem by revisiting the two key operations in Differentially Private Stochastic Gradient Descent (DP-SGD): 1) iterative perturbation and 2) gradient clipping. We propose a generic optimization framework, called {\\em ModelMix}, which performs random aggregation of intermediate model states. It strengthens the composite privacy analysis utilizing the entropy of the training trajectory and improves the $(\\epsilon, \\delta)$ DP security parameters ","authors_text":"Hanshen Xiao, Jun Wan, Srinivas Devadas","cross_cats":["cs.CR"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-10-07T22:59:00Z","title":"Differentially Private Deep Learning with ModelMix"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2210.03843","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:dbd829218f5036d8d9e06242d8c597319eecf4b2cc1b4526e5571cb34bec8d83","target":"record","created_at":"2026-07-05T05:04:32Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"8f25b1d6eb13070574344783a5607551120da9005d40dbd1969eb266c0fca83d","cross_cats_sorted":["cs.CR"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-10-07T22:59:00Z","title_canon_sha256":"57c3602b96eec67862c80a89003e96508e7036d79af2a2247fa1c5135a144ea8"},"schema_version":"1.0","source":{"id":"2210.03843","kind":"arxiv","version":1}},"canonical_sha256":"4057c057b80ca940b31fb89950ede5d5181d97ddf00ca799ac7ebe6b74b6ac74","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"4057c057b80ca940b31fb89950ede5d5181d97ddf00ca799ac7ebe6b74b6ac74","first_computed_at":"2026-07-05T05:04:32.922428Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T05:04:32.922428Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"z1xqrW5YcmPuIGoy3wiGycALOtAh4/pcNHEFZJG9d6iFACrXjtR03Iq5UCqjqmtU6ZDWi73VtkpeHiYlzBhFDA==","signature_status":"signed_v1","signed_at":"2026-07-05T05:04:32.922793Z","signed_message":"canonical_sha256_bytes"},"source_id":"2210.03843","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:dbd829218f5036d8d9e06242d8c597319eecf4b2cc1b4526e5571cb34bec8d83","sha256:ca7d07e53bb988095cf6423859f3d5cf9b58eb88769a3f8489c716e92577769e"],"state_sha256":"51a80d5ea55cbfda72da2df2a02519f4c83bf981ae99eb22044c83145f76594c"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"G2VxDu87qs+PboXxsPnUowujzrsMjNhZjbGLWqThIFvAF/8jbirxvA7KZzaRaaVkJBGkuVtHWSzO9+HEMnlUAw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-16T06:07:56.587549Z","bundle_sha256":"3b6bda047864ae00b94a56e6dfe77b725d26a4820cae1d8521ac05493c004ac3"}}