{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:2GTQP54PE5MBJV3SXTBV3MNILB","short_pith_number":"pith:2GTQP54P","schema_version":"1.0","canonical_sha256":"d1a707f78f275814d772bcc35db1a8586f100d19e493c6e9e34bc3a94e6ba1d1","source":{"kind":"arxiv","id":"2507.22398","version":3},"attestation_state":"computed","paper":{"title":"On the Reliability of Vision-Language Models Under Adversarial Frequency-Domain Perturbations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Ajmal Mian, Jordan Vice, Naveed Akhtar, Richard Hartley, Yansong Gao","submitted_at":"2025-07-30T05:41:29Z","abstract_excerpt":"Vision-Language Models (VLMs) are increasingly used as perceptual modules for visual content reasoning, including through captioning and DeepFake detection. In this work, we expose a critical vulnerability of VLMs when exposed to subtle, structured perturbations in the frequency domain. Specifically, we highlight how these feature transformations undermine authenticity/DeepFake detection and automated image captioning tasks. We design targeted image transformations, operating in the frequency domain to systematically adjust VLM outputs when exposed to frequency-perturbed real and synthetic ima"},"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":"2507.22398","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-07-30T05:41:29Z","cross_cats_sorted":[],"title_canon_sha256":"93425c2bfa20a3f09f01c083f920f898dee6aaee4fb2e5d8d1109f744b88f9dd","abstract_canon_sha256":"59a26ca20147ab0cbe82abe63e11fbd8caa3037fa6ce68edecf0a193118a0a29"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:53:06.228797Z","signature_b64":"r5ShGLvyxCVuhwLwZkE5ifzy1+s8t3YSS1y+2XVJenimnQJVKBZ54s+/DEV6b1XiRz4uzXj1DgloqnDxConaAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d1a707f78f275814d772bcc35db1a8586f100d19e493c6e9e34bc3a94e6ba1d1","last_reissued_at":"2026-07-05T11:53:06.228099Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:53:06.228099Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On the Reliability of Vision-Language Models Under Adversarial Frequency-Domain Perturbations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Ajmal Mian, Jordan Vice, Naveed Akhtar, Richard Hartley, Yansong Gao","submitted_at":"2025-07-30T05:41:29Z","abstract_excerpt":"Vision-Language Models (VLMs) are increasingly used as perceptual modules for visual content reasoning, including through captioning and DeepFake detection. In this work, we expose a critical vulnerability of VLMs when exposed to subtle, structured perturbations in the frequency domain. Specifically, we highlight how these feature transformations undermine authenticity/DeepFake detection and automated image captioning tasks. We design targeted image transformations, operating in the frequency domain to systematically adjust VLM outputs when exposed to frequency-perturbed real and synthetic ima"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.22398","kind":"arxiv","version":3},"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/2507.22398/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":"2507.22398","created_at":"2026-07-05T11:53:06.228178+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.22398v3","created_at":"2026-07-05T11:53:06.228178+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.22398","created_at":"2026-07-05T11:53:06.228178+00:00"},{"alias_kind":"pith_short_12","alias_value":"2GTQP54PE5MB","created_at":"2026-07-05T11:53:06.228178+00:00"},{"alias_kind":"pith_short_16","alias_value":"2GTQP54PE5MBJV3S","created_at":"2026-07-05T11:53:06.228178+00:00"},{"alias_kind":"pith_short_8","alias_value":"2GTQP54P","created_at":"2026-07-05T11:53:06.228178+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2GTQP54PE5MBJV3SXTBV3MNILB","json":"https://pith.science/pith/2GTQP54PE5MBJV3SXTBV3MNILB.json","graph_json":"https://pith.science/api/pith-number/2GTQP54PE5MBJV3SXTBV3MNILB/graph.json","events_json":"https://pith.science/api/pith-number/2GTQP54PE5MBJV3SXTBV3MNILB/events.json","paper":"https://pith.science/paper/2GTQP54P"},"agent_actions":{"view_html":"https://pith.science/pith/2GTQP54PE5MBJV3SXTBV3MNILB","download_json":"https://pith.science/pith/2GTQP54PE5MBJV3SXTBV3MNILB.json","view_paper":"https://pith.science/paper/2GTQP54P","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.22398&json=true","fetch_graph":"https://pith.science/api/pith-number/2GTQP54PE5MBJV3SXTBV3MNILB/graph.json","fetch_events":"https://pith.science/api/pith-number/2GTQP54PE5MBJV3SXTBV3MNILB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2GTQP54PE5MBJV3SXTBV3MNILB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2GTQP54PE5MBJV3SXTBV3MNILB/action/storage_attestation","attest_author":"https://pith.science/pith/2GTQP54PE5MBJV3SXTBV3MNILB/action/author_attestation","sign_citation":"https://pith.science/pith/2GTQP54PE5MBJV3SXTBV3MNILB/action/citation_signature","submit_replication":"https://pith.science/pith/2GTQP54PE5MBJV3SXTBV3MNILB/action/replication_record"}},"created_at":"2026-07-05T11:53:06.228178+00:00","updated_at":"2026-07-05T11:53:06.228178+00:00"}