{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:BXNI3C5MHM23GFOSBFQAYAHBUF","short_pith_number":"pith:BXNI3C5M","schema_version":"1.0","canonical_sha256":"0dda8d8bac3b35b315d209600c00e1a17df265cf588ca23a93aa6ca76ee62525","source":{"kind":"arxiv","id":"2507.05093","version":1},"attestation_state":"computed","paper":{"title":"The Hidden Threat in Plain Text: Attacking RAG Data Loaders","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Alberto Castagnaro, Luca Pajola, Mauro Conti, Simeone Pizzi, Umberto Salviati","submitted_at":"2025-07-07T15:13:54Z","abstract_excerpt":"Large Language Models (LLMs) have transformed human-machine interaction since ChatGPT's 2022 debut, with Retrieval-Augmented Generation (RAG) emerging as a key framework that enhances LLM outputs by integrating external knowledge. However, RAG's reliance on ingesting external documents introduces new vulnerabilities. This paper exposes a critical security gap at the data loading stage, where malicious actors can stealthily corrupt RAG pipelines by exploiting document ingestion.\n  We propose a taxonomy of 9 knowledge-based poisoning attacks and introduce two novel threat vectors -- Content Obfu"},"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":"2507.05093","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CR","submitted_at":"2025-07-07T15:13:54Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"fd572b1aa9b55dcd21b3046e6e404a9e06598b509b30372c0a0b956916081552","abstract_canon_sha256":"cf93ea8d4aa6c7b9aa8530d7cb50853c719d5b284d5ce8cb05fecdcf8d1a6ce5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:33:08.940048Z","signature_b64":"Htcp7tHT2fiBeibVJRq8aQ+bpdYK6X8Wc1/gC7sQbFljNSx5m7Ob/v9BV/ZeW3j72bu9FU5LQvdUd1HHAd57AA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0dda8d8bac3b35b315d209600c00e1a17df265cf588ca23a93aa6ca76ee62525","last_reissued_at":"2026-07-05T11:33:08.939639Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:33:08.939639Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Hidden Threat in Plain Text: Attacking RAG Data Loaders","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Alberto Castagnaro, Luca Pajola, Mauro Conti, Simeone Pizzi, Umberto Salviati","submitted_at":"2025-07-07T15:13:54Z","abstract_excerpt":"Large Language Models (LLMs) have transformed human-machine interaction since ChatGPT's 2022 debut, with Retrieval-Augmented Generation (RAG) emerging as a key framework that enhances LLM outputs by integrating external knowledge. However, RAG's reliance on ingesting external documents introduces new vulnerabilities. This paper exposes a critical security gap at the data loading stage, where malicious actors can stealthily corrupt RAG pipelines by exploiting document ingestion.\n  We propose a taxonomy of 9 knowledge-based poisoning attacks and introduce two novel threat vectors -- Content Obfu"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.05093","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/2507.05093/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":"2507.05093","created_at":"2026-07-05T11:33:08.939698+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.05093v1","created_at":"2026-07-05T11:33:08.939698+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.05093","created_at":"2026-07-05T11:33:08.939698+00:00"},{"alias_kind":"pith_short_12","alias_value":"BXNI3C5MHM23","created_at":"2026-07-05T11:33:08.939698+00:00"},{"alias_kind":"pith_short_16","alias_value":"BXNI3C5MHM23GFOS","created_at":"2026-07-05T11:33:08.939698+00:00"},{"alias_kind":"pith_short_8","alias_value":"BXNI3C5M","created_at":"2026-07-05T11:33:08.939698+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.00387","citing_title":"RAGShield: Detecting Numerical Claim Manipulation in Government RAG Systems","ref_index":10,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BXNI3C5MHM23GFOSBFQAYAHBUF","json":"https://pith.science/pith/BXNI3C5MHM23GFOSBFQAYAHBUF.json","graph_json":"https://pith.science/api/pith-number/BXNI3C5MHM23GFOSBFQAYAHBUF/graph.json","events_json":"https://pith.science/api/pith-number/BXNI3C5MHM23GFOSBFQAYAHBUF/events.json","paper":"https://pith.science/paper/BXNI3C5M"},"agent_actions":{"view_html":"https://pith.science/pith/BXNI3C5MHM23GFOSBFQAYAHBUF","download_json":"https://pith.science/pith/BXNI3C5MHM23GFOSBFQAYAHBUF.json","view_paper":"https://pith.science/paper/BXNI3C5M","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.05093&json=true","fetch_graph":"https://pith.science/api/pith-number/BXNI3C5MHM23GFOSBFQAYAHBUF/graph.json","fetch_events":"https://pith.science/api/pith-number/BXNI3C5MHM23GFOSBFQAYAHBUF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BXNI3C5MHM23GFOSBFQAYAHBUF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BXNI3C5MHM23GFOSBFQAYAHBUF/action/storage_attestation","attest_author":"https://pith.science/pith/BXNI3C5MHM23GFOSBFQAYAHBUF/action/author_attestation","sign_citation":"https://pith.science/pith/BXNI3C5MHM23GFOSBFQAYAHBUF/action/citation_signature","submit_replication":"https://pith.science/pith/BXNI3C5MHM23GFOSBFQAYAHBUF/action/replication_record"}},"created_at":"2026-07-05T11:33:08.939698+00:00","updated_at":"2026-07-05T11:33:08.939698+00:00"}