{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:W3CIG53NV52EWQKFUZ3NILAX6P","short_pith_number":"pith:W3CIG53N","schema_version":"1.0","canonical_sha256":"b6c483776daf744b4145a676d42c17f3d04ee79e55e05087f4236775c2e0e550","source":{"kind":"arxiv","id":"2503.05445","version":3},"attestation_state":"computed","paper":{"title":"Are Your LLM-based Text-to-SQL Models Secure? Exploring SQL Injection via Backdoor Attacks","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.DB"],"primary_cat":"cs.CR","authors_text":"Carl Yang, Haichuan Zhang, Jiale Lao, Meiyu Lin, Mingjie Tang, Renyuan Li, Yang Cao, Yuanchun Zhou","submitted_at":"2025-03-07T14:16:48Z","abstract_excerpt":"Large language models (LLMs) have shown state-of-the-art results in translating natural language questions into SQL queries (Text-to-SQL), a long-standing challenge within the database community. However, security concerns remain largely unexplored, particularly the threat of backdoor attacks, which can introduce malicious behaviors into models through fine-tuning with poisoned datasets. In this work, we systematically investigate the vulnerabilities of LLM-based Text-to-SQL models and present ToxicSQL, a novel backdoor attack framework. Our approach leverages stealthy {semantic and character-"},"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":"2503.05445","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CR","submitted_at":"2025-03-07T14:16:48Z","cross_cats_sorted":["cs.DB"],"title_canon_sha256":"fdb4aeb128b4dc824f7e8d19d7d15c75f222f58c86e7840f49e47ed163cd699a","abstract_canon_sha256":"08fa992b0c9af962fef8759ab6524d177d8585f51fbd2b7cf34ddaee2b82643a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:06:07.563996Z","signature_b64":"6QUUpqX/iI7Vwfx2haKrwLvnJIabj2eyf7fPmwgo4phLma03maXwkB6B/DUAsKIjbY9Po8KcklfUiwf9niwqAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b6c483776daf744b4145a676d42c17f3d04ee79e55e05087f4236775c2e0e550","last_reissued_at":"2026-07-05T12:06:07.563502Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:06:07.563502Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Are Your LLM-based Text-to-SQL Models Secure? Exploring SQL Injection via Backdoor Attacks","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.DB"],"primary_cat":"cs.CR","authors_text":"Carl Yang, Haichuan Zhang, Jiale Lao, Meiyu Lin, Mingjie Tang, Renyuan Li, Yang Cao, Yuanchun Zhou","submitted_at":"2025-03-07T14:16:48Z","abstract_excerpt":"Large language models (LLMs) have shown state-of-the-art results in translating natural language questions into SQL queries (Text-to-SQL), a long-standing challenge within the database community. However, security concerns remain largely unexplored, particularly the threat of backdoor attacks, which can introduce malicious behaviors into models through fine-tuning with poisoned datasets. In this work, we systematically investigate the vulnerabilities of LLM-based Text-to-SQL models and present ToxicSQL, a novel backdoor attack framework. Our approach leverages stealthy {semantic and character-"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.05445","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/2503.05445/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":"2503.05445","created_at":"2026-07-05T12:06:07.563558+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.05445v3","created_at":"2026-07-05T12:06:07.563558+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.05445","created_at":"2026-07-05T12:06:07.563558+00:00"},{"alias_kind":"pith_short_12","alias_value":"W3CIG53NV52E","created_at":"2026-07-05T12:06:07.563558+00:00"},{"alias_kind":"pith_short_16","alias_value":"W3CIG53NV52EWQKF","created_at":"2026-07-05T12:06:07.563558+00:00"},{"alias_kind":"pith_short_8","alias_value":"W3CIG53N","created_at":"2026-07-05T12:06:07.563558+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26627","citing_title":"Agents That Know Too Much: A Data-Centric Survey of Privacy in LLM Agents","ref_index":70,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/W3CIG53NV52EWQKFUZ3NILAX6P","json":"https://pith.science/pith/W3CIG53NV52EWQKFUZ3NILAX6P.json","graph_json":"https://pith.science/api/pith-number/W3CIG53NV52EWQKFUZ3NILAX6P/graph.json","events_json":"https://pith.science/api/pith-number/W3CIG53NV52EWQKFUZ3NILAX6P/events.json","paper":"https://pith.science/paper/W3CIG53N"},"agent_actions":{"view_html":"https://pith.science/pith/W3CIG53NV52EWQKFUZ3NILAX6P","download_json":"https://pith.science/pith/W3CIG53NV52EWQKFUZ3NILAX6P.json","view_paper":"https://pith.science/paper/W3CIG53N","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.05445&json=true","fetch_graph":"https://pith.science/api/pith-number/W3CIG53NV52EWQKFUZ3NILAX6P/graph.json","fetch_events":"https://pith.science/api/pith-number/W3CIG53NV52EWQKFUZ3NILAX6P/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/W3CIG53NV52EWQKFUZ3NILAX6P/action/timestamp_anchor","attest_storage":"https://pith.science/pith/W3CIG53NV52EWQKFUZ3NILAX6P/action/storage_attestation","attest_author":"https://pith.science/pith/W3CIG53NV52EWQKFUZ3NILAX6P/action/author_attestation","sign_citation":"https://pith.science/pith/W3CIG53NV52EWQKFUZ3NILAX6P/action/citation_signature","submit_replication":"https://pith.science/pith/W3CIG53NV52EWQKFUZ3NILAX6P/action/replication_record"}},"created_at":"2026-07-05T12:06:07.563558+00:00","updated_at":"2026-07-05T12:06:07.563558+00:00"}