{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:ZZFHQCTAF53N24HTPNL25NTTHK","short_pith_number":"pith:ZZFHQCTA","schema_version":"1.0","canonical_sha256":"ce4a780a602f76dd70f37b57aeb6733ab51226a9ab8fc78662b9c77a8c8dca09","source":{"kind":"arxiv","id":"2505.16640","version":1},"attestation_state":"computed","paper":{"title":"BadVLA: Towards Backdoor Attacks on Vision-Language-Action Models via Objective-Decoupled Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Guiyao Tie, Guowen Zhang, Hechang Wang, Lichao Sun, Pan Zhou, Xueyang Zhou","submitted_at":"2025-05-22T13:12:46Z","abstract_excerpt":"Vision-Language-Action (VLA) models have advanced robotic control by enabling end-to-end decision-making directly from multimodal inputs. However, their tightly coupled architectures expose novel security vulnerabilities. Unlike traditional adversarial perturbations, backdoor attacks represent a stealthier, persistent, and practically significant threat-particularly under the emerging Training-as-a-Service paradigm-but remain largely unexplored in the context of VLA models. To address this gap, we propose BadVLA, a backdoor attack method based on Objective-Decoupled Optimization, which for the"},"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":"2505.16640","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2025-05-22T13:12:46Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"757574fff497418aa974cdfdd2baa28bf17a1fd0a527de9898798b7c0b76402e","abstract_canon_sha256":"c228902f333be04c67388ac724683f8c3eef7fbfa6335f73e85eaaaa13d540b5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:07:38.466854Z","signature_b64":"OQPY41fR02cvsWHQsf44XTBEdkvPYxOElrKsjUGlih1ZZSPCo3s5bHrN+JxaalayVnqX6VVEMz7I2ABiqpJVAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ce4a780a602f76dd70f37b57aeb6733ab51226a9ab8fc78662b9c77a8c8dca09","last_reissued_at":"2026-07-05T11:07:38.466366Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:07:38.466366Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"BadVLA: Towards Backdoor Attacks on Vision-Language-Action Models via Objective-Decoupled Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Guiyao Tie, Guowen Zhang, Hechang Wang, Lichao Sun, Pan Zhou, Xueyang Zhou","submitted_at":"2025-05-22T13:12:46Z","abstract_excerpt":"Vision-Language-Action (VLA) models have advanced robotic control by enabling end-to-end decision-making directly from multimodal inputs. However, their tightly coupled architectures expose novel security vulnerabilities. Unlike traditional adversarial perturbations, backdoor attacks represent a stealthier, persistent, and practically significant threat-particularly under the emerging Training-as-a-Service paradigm-but remain largely unexplored in the context of VLA models. To address this gap, we propose BadVLA, a backdoor attack method based on Objective-Decoupled Optimization, which for the"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.16640","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/2505.16640/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":"2505.16640","created_at":"2026-07-05T11:07:38.466424+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.16640v1","created_at":"2026-07-05T11:07:38.466424+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.16640","created_at":"2026-07-05T11:07:38.466424+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZZFHQCTAF53N","created_at":"2026-07-05T11:07:38.466424+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZZFHQCTAF53N24HT","created_at":"2026-07-05T11:07:38.466424+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZZFHQCTA","created_at":"2026-07-05T11:07:38.466424+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":9,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.28083","citing_title":"VLA-Hijack: A Transferable Patch Attack against Vision-Language-Action Models via Visual Proprioception Hijacking","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2605.29114","citing_title":"ReasonBreak: Probing Vulnerabilities in Reasoning-Enabled Vision-Language-Action Models for Autonomous Driving","ref_index":55,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09499","citing_title":"Targeting World Models to Compromise Robot Learning Pipelines","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2602.03433","citing_title":"When control meets large language models: From words to dynamics","ref_index":209,"is_internal_anchor":false},{"citing_arxiv_id":"2603.24935","citing_title":"SABER: A Stealthy Agentic Black-Box Attack Framework for Vision-Language-Action Models","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2604.09651","citing_title":"FlowHijack: A Dynamics-Aware Backdoor Attack on Flow-Matching Vision-Language-Action Models","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09005","citing_title":"Towards Backdoor-Based Ownership Verification for Vision-Language-Action Models","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01950","citing_title":"TRAP: Tail-aware Ranking Attack for World-Model Planning","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07110","citing_title":"Securing Computer-Use Agents: A Unified Architecture-Lifecycle Framework for Deployment-Grounded Reliability","ref_index":157,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZZFHQCTAF53N24HTPNL25NTTHK","json":"https://pith.science/pith/ZZFHQCTAF53N24HTPNL25NTTHK.json","graph_json":"https://pith.science/api/pith-number/ZZFHQCTAF53N24HTPNL25NTTHK/graph.json","events_json":"https://pith.science/api/pith-number/ZZFHQCTAF53N24HTPNL25NTTHK/events.json","paper":"https://pith.science/paper/ZZFHQCTA"},"agent_actions":{"view_html":"https://pith.science/pith/ZZFHQCTAF53N24HTPNL25NTTHK","download_json":"https://pith.science/pith/ZZFHQCTAF53N24HTPNL25NTTHK.json","view_paper":"https://pith.science/paper/ZZFHQCTA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.16640&json=true","fetch_graph":"https://pith.science/api/pith-number/ZZFHQCTAF53N24HTPNL25NTTHK/graph.json","fetch_events":"https://pith.science/api/pith-number/ZZFHQCTAF53N24HTPNL25NTTHK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZZFHQCTAF53N24HTPNL25NTTHK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZZFHQCTAF53N24HTPNL25NTTHK/action/storage_attestation","attest_author":"https://pith.science/pith/ZZFHQCTAF53N24HTPNL25NTTHK/action/author_attestation","sign_citation":"https://pith.science/pith/ZZFHQCTAF53N24HTPNL25NTTHK/action/citation_signature","submit_replication":"https://pith.science/pith/ZZFHQCTAF53N24HTPNL25NTTHK/action/replication_record"}},"created_at":"2026-07-05T11:07:38.466424+00:00","updated_at":"2026-07-05T11:07:38.466424+00:00"}