{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:WNW5T5I5ZIWXTCBHMMUUK72S5H","short_pith_number":"pith:WNW5T5I5","schema_version":"1.0","canonical_sha256":"b36dd9f51dca2d7988276329457f52e9ecb332338dacfc988840c9a7cd3797aa","source":{"kind":"arxiv","id":"2109.12298","version":4},"attestation_state":"computed","paper":{"title":"Opacus: User-Friendly Differential Privacy Library in PyTorch","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.LG","authors_text":"Akash Bharadwaj, Alexandre Sablayrolles, Ashkan Yousefpour, Davide Testuggine, Graham Cormode, Igor Shilov, Ilya Mironov, Jessica Zhao, John Nguyen, Karthik Prasad, Mani Malek, Sayan Ghosh","submitted_at":"2021-09-25T07:10:54Z","abstract_excerpt":"We introduce Opacus, a free, open-source PyTorch library for training deep learning models with differential privacy (hosted at opacus.ai). Opacus is designed for simplicity, flexibility, and speed. It provides a simple and user-friendly API, and enables machine learning practitioners to make a training pipeline private by adding as little as two lines to their code. It supports a wide variety of layers, including multi-head attention, convolution, LSTM, GRU (and generic RNN), and embedding, right out of the box and provides the means for supporting other user-defined layers. Opacus computes b"},"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":"2109.12298","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-09-25T07:10:54Z","cross_cats_sorted":["cs.CR"],"title_canon_sha256":"fc0d92ed018eae14164a08e56b6e8ecfeaaa008e2babc02737b7e897cb36aea2","abstract_canon_sha256":"0e0c96a3f4c1784cf775171e11297c0656f5f50c3b076a8b514ff4bef2642208"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:50:31.984852Z","signature_b64":"a8oyPnOJj8vcoyrGPSzx6fRGydpybkr6brbu7F3ZdXDnplAlOBin/RkF1jfBSy+vFGNnZsgLnWOWFXXU3IDCCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b36dd9f51dca2d7988276329457f52e9ecb332338dacfc988840c9a7cd3797aa","last_reissued_at":"2026-07-05T04:50:31.984427Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:50:31.984427Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Opacus: User-Friendly Differential Privacy Library in PyTorch","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.LG","authors_text":"Akash Bharadwaj, Alexandre Sablayrolles, Ashkan Yousefpour, Davide Testuggine, Graham Cormode, Igor Shilov, Ilya Mironov, Jessica Zhao, John Nguyen, Karthik Prasad, Mani Malek, Sayan Ghosh","submitted_at":"2021-09-25T07:10:54Z","abstract_excerpt":"We introduce Opacus, a free, open-source PyTorch library for training deep learning models with differential privacy (hosted at opacus.ai). Opacus is designed for simplicity, flexibility, and speed. It provides a simple and user-friendly API, and enables machine learning practitioners to make a training pipeline private by adding as little as two lines to their code. It supports a wide variety of layers, including multi-head attention, convolution, LSTM, GRU (and generic RNN), and embedding, right out of the box and provides the means for supporting other user-defined layers. Opacus computes b"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2109.12298","kind":"arxiv","version":4},"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/2109.12298/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":"2109.12298","created_at":"2026-07-05T04:50:31.984486+00:00"},{"alias_kind":"arxiv_version","alias_value":"2109.12298v4","created_at":"2026-07-05T04:50:31.984486+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2109.12298","created_at":"2026-07-05T04:50:31.984486+00:00"},{"alias_kind":"pith_short_12","alias_value":"WNW5T5I5ZIWX","created_at":"2026-07-05T04:50:31.984486+00:00"},{"alias_kind":"pith_short_16","alias_value":"WNW5T5I5ZIWXTCBH","created_at":"2026-07-05T04:50:31.984486+00:00"},{"alias_kind":"pith_short_8","alias_value":"WNW5T5I5","created_at":"2026-07-05T04:50:31.984486+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":27,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.06320","citing_title":"Dithered Gaussian Mechanism for Randomness-Efficient Differential Privacy","ref_index":51,"is_internal_anchor":true},{"citing_arxiv_id":"2606.11556","citing_title":"Privacy-Preserving Federated Autoencoder for ECG Anomaly Detection on Edge Devices","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2606.01719","citing_title":"Fair Finetuning Mitigates Distribution Inference Attacks","ref_index":68,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06505","citing_title":"PACZero: PAC-Private Fine-Tuning of Language Models via Sign Quantization","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17373","citing_title":"FML-bench: A Controlled Study of AI Research Agent Strategies from the Perspective of Search Dynamics","ref_index":59,"is_internal_anchor":false},{"citing_arxiv_id":"2605.24879","citing_title":"Efficient DP-SGD for LLMs with Randomized Clipping","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2605.26243","citing_title":"Provably Communication-Efficient and Privacy-Preserving Federated Graph Neural Networks","ref_index":86,"is_internal_anchor":false},{"citing_arxiv_id":"2411.04680","citing_title":"Privacy Leakage via Output Label Space and Differentially Private Continual Learning","ref_index":84,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21780","citing_title":"Provable Robustness against Backdoor Attacks via the Primal-Dual Perspective on Differential Privacy","ref_index":77,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20521","citing_title":"An exponential mechanism based on quadratic approximations for fine-tuning machine learning models with privacy guarantees","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15851","citing_title":"DPrivBench: Benchmarking LLMs' Reasoning for Differential Privacy","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17373","citing_title":"FML-bench: A Controlled Study of AI Research Agent Strategies from the Perspective of Search Dynamics","ref_index":59,"is_internal_anchor":false},{"citing_arxiv_id":"2509.03472","citing_title":"DPQuant: Efficient and Differentially-Private Model Training via Dynamic Quantization Scheduling","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2510.20616","citing_title":"On Optimal Hyperparameters for Differentially Private Deep Transfer Learning","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2511.21804","citing_title":"Beyond Membership: Limitations of Add/Remove Adjacency in Differential Privacy","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2512.18809","citing_title":"FedVideoMAE: Efficient Privacy-Preserving Federated Video Moderation","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2601.10237","citing_title":"Fundamental Limitations of Favorable Privacy-Utility Guarantees for DP-SGD","ref_index":65,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03425","citing_title":"FIBER: A Differentially Private Optimizer with Filter-Aware Innovation Bias Correction","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2604.26467","citing_title":"Differentially Private Contrastive Learning via Bounding Group-level Contribution","ref_index":54,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02718","citing_title":"Private Speech Classification without Collapse: Stabilized DP Training and Offline Distillation","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06505","citing_title":"PACZero: PAC-Private Fine-Tuning of Language Models via Sign Quantization","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05769","citing_title":"Adaptive Selection of LoRA Components in Privacy-Preserving Federated Learning","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20985","citing_title":"Differentially Private Model Merging","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01679","citing_title":"Class-Aware Adaptive Differential Privacy in Deep Learning for Sensor-Based Fall Detection","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07493","citing_title":"Differentially Private Modeling of Disease Transmission within Human Contact Networks","ref_index":104,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WNW5T5I5ZIWXTCBHMMUUK72S5H","json":"https://pith.science/pith/WNW5T5I5ZIWXTCBHMMUUK72S5H.json","graph_json":"https://pith.science/api/pith-number/WNW5T5I5ZIWXTCBHMMUUK72S5H/graph.json","events_json":"https://pith.science/api/pith-number/WNW5T5I5ZIWXTCBHMMUUK72S5H/events.json","paper":"https://pith.science/paper/WNW5T5I5"},"agent_actions":{"view_html":"https://pith.science/pith/WNW5T5I5ZIWXTCBHMMUUK72S5H","download_json":"https://pith.science/pith/WNW5T5I5ZIWXTCBHMMUUK72S5H.json","view_paper":"https://pith.science/paper/WNW5T5I5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2109.12298&json=true","fetch_graph":"https://pith.science/api/pith-number/WNW5T5I5ZIWXTCBHMMUUK72S5H/graph.json","fetch_events":"https://pith.science/api/pith-number/WNW5T5I5ZIWXTCBHMMUUK72S5H/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WNW5T5I5ZIWXTCBHMMUUK72S5H/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WNW5T5I5ZIWXTCBHMMUUK72S5H/action/storage_attestation","attest_author":"https://pith.science/pith/WNW5T5I5ZIWXTCBHMMUUK72S5H/action/author_attestation","sign_citation":"https://pith.science/pith/WNW5T5I5ZIWXTCBHMMUUK72S5H/action/citation_signature","submit_replication":"https://pith.science/pith/WNW5T5I5ZIWXTCBHMMUUK72S5H/action/replication_record"}},"created_at":"2026-07-05T04:50:31.984486+00:00","updated_at":"2026-07-05T04:50:31.984486+00:00"}