{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:MUNH5FD3JLRQZKP5O7ERQFVCB3","short_pith_number":"pith:MUNH5FD3","schema_version":"1.0","canonical_sha256":"651a7e947b4ae30ca9fd77c91816a20eda0df16f2f96b14b0fea0345f775ed8f","source":{"kind":"arxiv","id":"2404.13518","version":1},"attestation_state":"computed","paper":{"title":"Reliable Model Watermarking: Defending Against Theft without Compromising on Evasion","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Fangqi Li, Hongyu Zhu, Ju Jia, Shilin Wang, Sichu Liang, Wentao Hu","submitted_at":"2024-04-21T03:38:20Z","abstract_excerpt":"With the rise of Machine Learning as a Service (MLaaS) platforms,safeguarding the intellectual property of deep learning models is becoming paramount. Among various protective measures, trigger set watermarking has emerged as a flexible and effective strategy for preventing unauthorized model distribution. However, this paper identifies an inherent flaw in the current paradigm of trigger set watermarking: evasion adversaries can readily exploit the shortcuts created by models memorizing watermark samples that deviate from the main task distribution, significantly impairing their generalization"},"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":"2404.13518","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2024-04-21T03:38:20Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"7bacd37bcd1c398847541fbe1fb95452f44a6b44aca0a16a591d66632dcb6fde","abstract_canon_sha256":"65668f517170d1b5235c860621ae111623bcd903c6e0ea2b7fd5b79ec0052afa"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:10:33.506797Z","signature_b64":"ILYtOrslqzYBL+Whys3SHgz5tmFsvQHjHAHjeIc3DjJCBocxKJ6XKDWrk24wkp8mbLimndulLtLi0qK515W5Cw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"651a7e947b4ae30ca9fd77c91816a20eda0df16f2f96b14b0fea0345f775ed8f","last_reissued_at":"2026-07-05T08:10:33.506339Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:10:33.506339Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Reliable Model Watermarking: Defending Against Theft without Compromising on Evasion","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Fangqi Li, Hongyu Zhu, Ju Jia, Shilin Wang, Sichu Liang, Wentao Hu","submitted_at":"2024-04-21T03:38:20Z","abstract_excerpt":"With the rise of Machine Learning as a Service (MLaaS) platforms,safeguarding the intellectual property of deep learning models is becoming paramount. Among various protective measures, trigger set watermarking has emerged as a flexible and effective strategy for preventing unauthorized model distribution. However, this paper identifies an inherent flaw in the current paradigm of trigger set watermarking: evasion adversaries can readily exploit the shortcuts created by models memorizing watermark samples that deviate from the main task distribution, significantly impairing their generalization"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.13518","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/2404.13518/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":"2404.13518","created_at":"2026-07-05T08:10:33.506400+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.13518v1","created_at":"2026-07-05T08:10:33.506400+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.13518","created_at":"2026-07-05T08:10:33.506400+00:00"},{"alias_kind":"pith_short_12","alias_value":"MUNH5FD3JLRQ","created_at":"2026-07-05T08:10:33.506400+00:00"},{"alias_kind":"pith_short_16","alias_value":"MUNH5FD3JLRQZKP5","created_at":"2026-07-05T08:10:33.506400+00:00"},{"alias_kind":"pith_short_8","alias_value":"MUNH5FD3","created_at":"2026-07-05T08:10:33.506400+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2508.15031","citing_title":"A Systematic Survey of Model Extraction Attacks and Defenses: State-of-the-Art and Perspectives","ref_index":277,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MUNH5FD3JLRQZKP5O7ERQFVCB3","json":"https://pith.science/pith/MUNH5FD3JLRQZKP5O7ERQFVCB3.json","graph_json":"https://pith.science/api/pith-number/MUNH5FD3JLRQZKP5O7ERQFVCB3/graph.json","events_json":"https://pith.science/api/pith-number/MUNH5FD3JLRQZKP5O7ERQFVCB3/events.json","paper":"https://pith.science/paper/MUNH5FD3"},"agent_actions":{"view_html":"https://pith.science/pith/MUNH5FD3JLRQZKP5O7ERQFVCB3","download_json":"https://pith.science/pith/MUNH5FD3JLRQZKP5O7ERQFVCB3.json","view_paper":"https://pith.science/paper/MUNH5FD3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.13518&json=true","fetch_graph":"https://pith.science/api/pith-number/MUNH5FD3JLRQZKP5O7ERQFVCB3/graph.json","fetch_events":"https://pith.science/api/pith-number/MUNH5FD3JLRQZKP5O7ERQFVCB3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MUNH5FD3JLRQZKP5O7ERQFVCB3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MUNH5FD3JLRQZKP5O7ERQFVCB3/action/storage_attestation","attest_author":"https://pith.science/pith/MUNH5FD3JLRQZKP5O7ERQFVCB3/action/author_attestation","sign_citation":"https://pith.science/pith/MUNH5FD3JLRQZKP5O7ERQFVCB3/action/citation_signature","submit_replication":"https://pith.science/pith/MUNH5FD3JLRQZKP5O7ERQFVCB3/action/replication_record"}},"created_at":"2026-07-05T08:10:33.506400+00:00","updated_at":"2026-07-05T08:10:33.506400+00:00"}