{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:METYTBWAKP3QORWS3T3XIF3OFQ","short_pith_number":"pith:METYTBWA","schema_version":"1.0","canonical_sha256":"61278986c053f70746d2dcf774176e2c3ce229aa5531bd83be3000c3259ed1be","source":{"kind":"arxiv","id":"2411.18191","version":2},"attestation_state":"computed","paper":{"title":"InputSnatch: Stealing Input in LLM Services via Timing Side-Channel Attacks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CR","authors_text":"Husheng Han, Qi Guo, Qiyan Fang, Shangyi Shi, Xing Hu, Xinyao Zheng, Zidong Du","submitted_at":"2024-11-27T10:14:38Z","abstract_excerpt":"Large language models (LLMs) possess extensive knowledge and question-answering capabilities, having been widely deployed in privacy-sensitive domains like finance and medical consultation. During LLM inferences, cache-sharing methods are commonly employed to enhance efficiency by reusing cached states or responses for the same or similar inference requests. However, we identify that these cache mechanisms pose a risk of private input leakage, as the caching can result in observable variations in response times, making them a strong candidate for a timing-based attack hint.\n  In this study, we"},"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":"2411.18191","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2024-11-27T10:14:38Z","cross_cats_sorted":[],"title_canon_sha256":"27c09f2b82b915f6b05515e94329ea4069ced8a2abdde2bebb0b98886fbaaf47","abstract_canon_sha256":"1f492758e474fd42e892a5dad8d03c373f32c8574d70ee9234d800588a00f32c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:41:47.351138Z","signature_b64":"8GaL4hjZSiJtdXBc9cbQfAD/SDXpt7QbONqJbKQ3FlRzfAWOsKcwxmOIfPksRR1aMAZ+RqmozC3Qr4cLXxpgBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"61278986c053f70746d2dcf774176e2c3ce229aa5531bd83be3000c3259ed1be","last_reissued_at":"2026-07-05T09:41:47.349192Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:41:47.349192Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"InputSnatch: Stealing Input in LLM Services via Timing Side-Channel Attacks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CR","authors_text":"Husheng Han, Qi Guo, Qiyan Fang, Shangyi Shi, Xing Hu, Xinyao Zheng, Zidong Du","submitted_at":"2024-11-27T10:14:38Z","abstract_excerpt":"Large language models (LLMs) possess extensive knowledge and question-answering capabilities, having been widely deployed in privacy-sensitive domains like finance and medical consultation. During LLM inferences, cache-sharing methods are commonly employed to enhance efficiency by reusing cached states or responses for the same or similar inference requests. However, we identify that these cache mechanisms pose a risk of private input leakage, as the caching can result in observable variations in response times, making them a strong candidate for a timing-based attack hint.\n  In this study, we"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.18191","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/2411.18191/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":"2411.18191","created_at":"2026-07-05T09:41:47.350519+00:00"},{"alias_kind":"arxiv_version","alias_value":"2411.18191v2","created_at":"2026-07-05T09:41:47.350519+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.18191","created_at":"2026-07-05T09:41:47.350519+00:00"},{"alias_kind":"pith_short_12","alias_value":"METYTBWAKP3Q","created_at":"2026-07-05T09:41:47.350519+00:00"},{"alias_kind":"pith_short_16","alias_value":"METYTBWAKP3QORWS","created_at":"2026-07-05T09:41:47.350519+00:00"},{"alias_kind":"pith_short_8","alias_value":"METYTBWA","created_at":"2026-07-05T09:41:47.350519+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.22237","citing_title":"Investigating The Security of Modern AI and Cloud Infrastructure","ref_index":151,"is_internal_anchor":false},{"citing_arxiv_id":"2605.23640","citing_title":"CachePrune: Privacy-Aware and Fine-Grained KV Cache Sharing for Efficient LLM Inference","ref_index":68,"is_internal_anchor":false},{"citing_arxiv_id":"2603.10726","citing_title":"PrefixWall: Mitigating Prefix Caching Side Channels in Shared LLM Systems","ref_index":76,"is_internal_anchor":false},{"citing_arxiv_id":"2603.09002","citing_title":"Security Considerations for Multi-agent Systems","ref_index":277,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/METYTBWAKP3QORWS3T3XIF3OFQ","json":"https://pith.science/pith/METYTBWAKP3QORWS3T3XIF3OFQ.json","graph_json":"https://pith.science/api/pith-number/METYTBWAKP3QORWS3T3XIF3OFQ/graph.json","events_json":"https://pith.science/api/pith-number/METYTBWAKP3QORWS3T3XIF3OFQ/events.json","paper":"https://pith.science/paper/METYTBWA"},"agent_actions":{"view_html":"https://pith.science/pith/METYTBWAKP3QORWS3T3XIF3OFQ","download_json":"https://pith.science/pith/METYTBWAKP3QORWS3T3XIF3OFQ.json","view_paper":"https://pith.science/paper/METYTBWA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2411.18191&json=true","fetch_graph":"https://pith.science/api/pith-number/METYTBWAKP3QORWS3T3XIF3OFQ/graph.json","fetch_events":"https://pith.science/api/pith-number/METYTBWAKP3QORWS3T3XIF3OFQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/METYTBWAKP3QORWS3T3XIF3OFQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/METYTBWAKP3QORWS3T3XIF3OFQ/action/storage_attestation","attest_author":"https://pith.science/pith/METYTBWAKP3QORWS3T3XIF3OFQ/action/author_attestation","sign_citation":"https://pith.science/pith/METYTBWAKP3QORWS3T3XIF3OFQ/action/citation_signature","submit_replication":"https://pith.science/pith/METYTBWAKP3QORWS3T3XIF3OFQ/action/replication_record"}},"created_at":"2026-07-05T09:41:47.350519+00:00","updated_at":"2026-07-05T09:41:47.350519+00:00"}