{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:5F7ANLJHBYBDLBEZZIFXS2CMAV","short_pith_number":"pith:5F7ANLJH","schema_version":"1.0","canonical_sha256":"e97e06ad270e02358499ca0b79684c0561f0dd4ce3f0de15be730f16198fdb72","source":{"kind":"arxiv","id":"2403.11981","version":2},"attestation_state":"computed","paper":{"title":"Certified Robustness to Clean-Label Poisoning Using Diffusion Denoising","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","cs.LG"],"primary_cat":"cs.CR","authors_text":"Alexey Kurakin, Nicholas Carlini, Sanghyun Hong","submitted_at":"2024-03-18T17:17:07Z","abstract_excerpt":"We present a certified defense to clean-label poisoning attacks under $\\ell_2$-norm. These attacks work by injecting a small number of poisoning samples (e.g., 1%) that contain bounded adversarial perturbations into the training data to induce a targeted misclassification of a test-time input. Inspired by the adversarial robustness achieved by $randomized$ $smoothing$, we show how an off-the-shelf diffusion denoising model can sanitize the tampered training data. We extensively test our defense against seven clean-label poisoning attacks in both $\\ell_2$ and $\\ell_{\\infty}$-norms and reduce th"},"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":"2403.11981","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2024-03-18T17:17:07Z","cross_cats_sorted":["cs.CV","cs.LG"],"title_canon_sha256":"9020ba75b6a99e4e9de97b60aba1609ba36cd20c3129ba4f0a940777fd57270d","abstract_canon_sha256":"312e4b38a3a6a89f341d4511c2e3dfa4f930fe471f66293e3e90091f40aca4a5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:13:55.520946Z","signature_b64":"AyUkeCcZHOCWPl1ldyslOqjWp51VpWcc2nVL2lDek8mOSL74AkJNT3eZcrFtUghEj/5pjA7sDf9QmUEd76E2Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e97e06ad270e02358499ca0b79684c0561f0dd4ce3f0de15be730f16198fdb72","last_reissued_at":"2026-07-05T11:13:55.520441Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:13:55.520441Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Certified Robustness to Clean-Label Poisoning Using Diffusion Denoising","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","cs.LG"],"primary_cat":"cs.CR","authors_text":"Alexey Kurakin, Nicholas Carlini, Sanghyun Hong","submitted_at":"2024-03-18T17:17:07Z","abstract_excerpt":"We present a certified defense to clean-label poisoning attacks under $\\ell_2$-norm. These attacks work by injecting a small number of poisoning samples (e.g., 1%) that contain bounded adversarial perturbations into the training data to induce a targeted misclassification of a test-time input. Inspired by the adversarial robustness achieved by $randomized$ $smoothing$, we show how an off-the-shelf diffusion denoising model can sanitize the tampered training data. We extensively test our defense against seven clean-label poisoning attacks in both $\\ell_2$ and $\\ell_{\\infty}$-norms and reduce th"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.11981","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/2403.11981/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":"2403.11981","created_at":"2026-07-05T11:13:55.520502+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.11981v2","created_at":"2026-07-05T11:13:55.520502+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.11981","created_at":"2026-07-05T11:13:55.520502+00:00"},{"alias_kind":"pith_short_12","alias_value":"5F7ANLJHBYBD","created_at":"2026-07-05T11:13:55.520502+00:00"},{"alias_kind":"pith_short_16","alias_value":"5F7ANLJHBYBDLBEZ","created_at":"2026-07-05T11:13:55.520502+00:00"},{"alias_kind":"pith_short_8","alias_value":"5F7ANLJH","created_at":"2026-07-05T11:13:55.520502+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2412.01528","citing_title":"CopyrightShield: Enhancing Diffusion Model Security against Copyright Infringement Attacks","ref_index":12,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5F7ANLJHBYBDLBEZZIFXS2CMAV","json":"https://pith.science/pith/5F7ANLJHBYBDLBEZZIFXS2CMAV.json","graph_json":"https://pith.science/api/pith-number/5F7ANLJHBYBDLBEZZIFXS2CMAV/graph.json","events_json":"https://pith.science/api/pith-number/5F7ANLJHBYBDLBEZZIFXS2CMAV/events.json","paper":"https://pith.science/paper/5F7ANLJH"},"agent_actions":{"view_html":"https://pith.science/pith/5F7ANLJHBYBDLBEZZIFXS2CMAV","download_json":"https://pith.science/pith/5F7ANLJHBYBDLBEZZIFXS2CMAV.json","view_paper":"https://pith.science/paper/5F7ANLJH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.11981&json=true","fetch_graph":"https://pith.science/api/pith-number/5F7ANLJHBYBDLBEZZIFXS2CMAV/graph.json","fetch_events":"https://pith.science/api/pith-number/5F7ANLJHBYBDLBEZZIFXS2CMAV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5F7ANLJHBYBDLBEZZIFXS2CMAV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5F7ANLJHBYBDLBEZZIFXS2CMAV/action/storage_attestation","attest_author":"https://pith.science/pith/5F7ANLJHBYBDLBEZZIFXS2CMAV/action/author_attestation","sign_citation":"https://pith.science/pith/5F7ANLJHBYBDLBEZZIFXS2CMAV/action/citation_signature","submit_replication":"https://pith.science/pith/5F7ANLJHBYBDLBEZZIFXS2CMAV/action/replication_record"}},"created_at":"2026-07-05T11:13:55.520502+00:00","updated_at":"2026-07-05T11:13:55.520502+00:00"}