{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:ZNKYOHG2WER4QN6U6WXKHZEPV6","short_pith_number":"pith:ZNKYOHG2","schema_version":"1.0","canonical_sha256":"cb55871cdab123c837d4f5aea3e48fafbd10a176718a13d67ad6badbade0fe4e","source":{"kind":"arxiv","id":"2606.28953","version":1},"attestation_state":"computed","paper":{"title":"Clustering Unsupervised Representations as Defense against Poisoning Attacks on Speech Commands Classification System","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.SD","authors_text":"Henry Li, Jesus Villalba, Martin Sustek, Najim Dehak, Sanjeev Khudanpur, Sonal Joshi, Thomas Thebaud","submitted_at":"2026-06-27T14:51:30Z","abstract_excerpt":"Poisoning attacks entail attackers intentionally tampering with training data. In this paper, we consider a dirty-label poisoning attack scenario on a speech commands classification system. The threat model assumes that certain utterances from one of the classes (source class) are poisoned by superimposing a trigger on it, and its label is changed to another class selected by the attacker (target class). We propose a filtering defense against such an attack. First, we use DIstillation with NO labels (DINO) to learn unsupervised representations for all the training examples. Next, we use K-mean"},"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":"2606.28953","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.SD","submitted_at":"2026-06-27T14:51:30Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"36af3110de7990d9b0afc8f57d8a6264f7c072c36370a2457ab75d58df7ec300","abstract_canon_sha256":"4a6f5963e77f537a198f2fb7d28b883b27f6b3fb3f845877445c7764006f10aa"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-30T01:17:47.045527Z","signature_b64":"skD57NOPPDwThYXsggAIYDgBoGb5qqPLZMKutqBHlYkb/LX3sAx8KZyZmvvAP4DtOIU8n28kglV3+l8zmmlXBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cb55871cdab123c837d4f5aea3e48fafbd10a176718a13d67ad6badbade0fe4e","last_reissued_at":"2026-06-30T01:17:47.044926Z","signature_status":"signed_v1","first_computed_at":"2026-06-30T01:17:47.044926Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Clustering Unsupervised Representations as Defense against Poisoning Attacks on Speech Commands Classification System","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.SD","authors_text":"Henry Li, Jesus Villalba, Martin Sustek, Najim Dehak, Sanjeev Khudanpur, Sonal Joshi, Thomas Thebaud","submitted_at":"2026-06-27T14:51:30Z","abstract_excerpt":"Poisoning attacks entail attackers intentionally tampering with training data. In this paper, we consider a dirty-label poisoning attack scenario on a speech commands classification system. The threat model assumes that certain utterances from one of the classes (source class) are poisoned by superimposing a trigger on it, and its label is changed to another class selected by the attacker (target class). We propose a filtering defense against such an attack. First, we use DIstillation with NO labels (DINO) to learn unsupervised representations for all the training examples. Next, we use K-mean"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.28953","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/2606.28953/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":"2606.28953","created_at":"2026-06-30T01:17:47.045010+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.28953v1","created_at":"2026-06-30T01:17:47.045010+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.28953","created_at":"2026-06-30T01:17:47.045010+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZNKYOHG2WER4","created_at":"2026-06-30T01:17:47.045010+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZNKYOHG2WER4QN6U","created_at":"2026-06-30T01:17:47.045010+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZNKYOHG2","created_at":"2026-06-30T01:17:47.045010+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2606.28953","citing_title":"Clustering Unsupervised Representations as Defense against Poisoning Attacks on Speech Commands Classification System","ref_index":1,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZNKYOHG2WER4QN6U6WXKHZEPV6","json":"https://pith.science/pith/ZNKYOHG2WER4QN6U6WXKHZEPV6.json","graph_json":"https://pith.science/api/pith-number/ZNKYOHG2WER4QN6U6WXKHZEPV6/graph.json","events_json":"https://pith.science/api/pith-number/ZNKYOHG2WER4QN6U6WXKHZEPV6/events.json","paper":"https://pith.science/paper/ZNKYOHG2"},"agent_actions":{"view_html":"https://pith.science/pith/ZNKYOHG2WER4QN6U6WXKHZEPV6","download_json":"https://pith.science/pith/ZNKYOHG2WER4QN6U6WXKHZEPV6.json","view_paper":"https://pith.science/paper/ZNKYOHG2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.28953&json=true","fetch_graph":"https://pith.science/api/pith-number/ZNKYOHG2WER4QN6U6WXKHZEPV6/graph.json","fetch_events":"https://pith.science/api/pith-number/ZNKYOHG2WER4QN6U6WXKHZEPV6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZNKYOHG2WER4QN6U6WXKHZEPV6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZNKYOHG2WER4QN6U6WXKHZEPV6/action/storage_attestation","attest_author":"https://pith.science/pith/ZNKYOHG2WER4QN6U6WXKHZEPV6/action/author_attestation","sign_citation":"https://pith.science/pith/ZNKYOHG2WER4QN6U6WXKHZEPV6/action/citation_signature","submit_replication":"https://pith.science/pith/ZNKYOHG2WER4QN6U6WXKHZEPV6/action/replication_record"}},"created_at":"2026-06-30T01:17:47.045010+00:00","updated_at":"2026-06-30T01:17:47.045010+00:00"}