{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:3TQXPMPEFPCRFIU5RYW7XARET4","short_pith_number":"pith:3TQXPMPE","schema_version":"1.0","canonical_sha256":"dce177b1e42bc512a29d8e2dfb82249f093adec8927e8a228cf5a4b67287360e","source":{"kind":"arxiv","id":"2603.28378","version":2},"attestation_state":"computed","paper":{"title":"Membership Inference Attacks against Large Audio Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.SD","authors_text":"Hung-yi Lee, Jia-Kai Dong, Yu-Xiang Lin","submitted_at":"2026-03-30T12:45:28Z","abstract_excerpt":"We present the first systematic Membership Inference Attack (MIA) evaluation of LALMs. Using Multi-modal Blind Baselines based on textual, spectral and prosodic features, we demonstrate that common audio datasets exhibit near-perfect train/test separability (AUC ~ 1.0) even without model inference, thus MIA may primarily detect distribution shift. We therefore introduce a blind-baseline protocol to control for this confound. Under this protocol, we identify that the distribution-matched datasets enable reliable MIA evaluation without distribution-shift artifacts. We benchmark multiple MIA meth"},"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":"2603.28378","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SD","submitted_at":"2026-03-30T12:45:28Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"5c91b9afacf124792339e84b185337ccee36df3de4ea44d1b095030801d71679","abstract_canon_sha256":"1c18fff7fdb348c1abe74b4bca04930147bb20a84b1f24897699868f0adffddb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-19T16:09:57.606626Z","signature_b64":"BhcwGih5g2ui0o5LZI9NeNH39vWsK5Wu/p49uJZ++rtZYSurGCLkOta02e5LbaBryGnbIwNDFTnx0+PclCeSAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dce177b1e42bc512a29d8e2dfb82249f093adec8927e8a228cf5a4b67287360e","last_reissued_at":"2026-06-19T16:09:57.606224Z","signature_status":"signed_v1","first_computed_at":"2026-06-19T16:09:57.606224Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Membership Inference Attacks against Large Audio Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.SD","authors_text":"Hung-yi Lee, Jia-Kai Dong, Yu-Xiang Lin","submitted_at":"2026-03-30T12:45:28Z","abstract_excerpt":"We present the first systematic Membership Inference Attack (MIA) evaluation of LALMs. Using Multi-modal Blind Baselines based on textual, spectral and prosodic features, we demonstrate that common audio datasets exhibit near-perfect train/test separability (AUC ~ 1.0) even without model inference, thus MIA may primarily detect distribution shift. We therefore introduce a blind-baseline protocol to control for this confound. Under this protocol, we identify that the distribution-matched datasets enable reliable MIA evaluation without distribution-shift artifacts. We benchmark multiple MIA meth"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.28378","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/2603.28378/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":"2603.28378","created_at":"2026-06-19T16:09:57.606279+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.28378v2","created_at":"2026-06-19T16:09:57.606279+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.28378","created_at":"2026-06-19T16:09:57.606279+00:00"},{"alias_kind":"pith_short_12","alias_value":"3TQXPMPEFPCR","created_at":"2026-06-19T16:09:57.606279+00:00"},{"alias_kind":"pith_short_16","alias_value":"3TQXPMPEFPCRFIU5","created_at":"2026-06-19T16:09:57.606279+00:00"},{"alias_kind":"pith_short_8","alias_value":"3TQXPMPE","created_at":"2026-06-19T16:09:57.606279+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2604.25591","citing_title":"Walking Through Uncertainty: An Empirical Study of Uncertainty Estimation for Audio-Aware Large Language Models","ref_index":47,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3TQXPMPEFPCRFIU5RYW7XARET4","json":"https://pith.science/pith/3TQXPMPEFPCRFIU5RYW7XARET4.json","graph_json":"https://pith.science/api/pith-number/3TQXPMPEFPCRFIU5RYW7XARET4/graph.json","events_json":"https://pith.science/api/pith-number/3TQXPMPEFPCRFIU5RYW7XARET4/events.json","paper":"https://pith.science/paper/3TQXPMPE"},"agent_actions":{"view_html":"https://pith.science/pith/3TQXPMPEFPCRFIU5RYW7XARET4","download_json":"https://pith.science/pith/3TQXPMPEFPCRFIU5RYW7XARET4.json","view_paper":"https://pith.science/paper/3TQXPMPE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.28378&json=true","fetch_graph":"https://pith.science/api/pith-number/3TQXPMPEFPCRFIU5RYW7XARET4/graph.json","fetch_events":"https://pith.science/api/pith-number/3TQXPMPEFPCRFIU5RYW7XARET4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3TQXPMPEFPCRFIU5RYW7XARET4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3TQXPMPEFPCRFIU5RYW7XARET4/action/storage_attestation","attest_author":"https://pith.science/pith/3TQXPMPEFPCRFIU5RYW7XARET4/action/author_attestation","sign_citation":"https://pith.science/pith/3TQXPMPEFPCRFIU5RYW7XARET4/action/citation_signature","submit_replication":"https://pith.science/pith/3TQXPMPEFPCRFIU5RYW7XARET4/action/replication_record"}},"created_at":"2026-06-19T16:09:57.606279+00:00","updated_at":"2026-06-19T16:09:57.606279+00:00"}