{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:7BJ3PWTSLFBUYKWN7KHIKYCHHT","short_pith_number":"pith:7BJ3PWTS","schema_version":"1.0","canonical_sha256":"f853b7da7259434c2acdfa8e8560473cece8db365e99d9cb96d49acd5b3fe03d","source":{"kind":"arxiv","id":"2404.16251","version":3},"attestation_state":"computed","paper":{"title":"Prompt Leakage effect and defense strategies for multi-turn LLM interactions","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.CR","authors_text":"Alexander R. Fabbri, Ben Risher, Chien-Sheng Wu, Divyansh Agarwal, Philippe Laban, Shafiq Joty","submitted_at":"2024-04-24T23:39:58Z","abstract_excerpt":"Prompt leakage poses a compelling security and privacy threat in LLM applications. Leakage of system prompts may compromise intellectual property, and act as adversarial reconnaissance for an attacker. A systematic evaluation of prompt leakage threats and mitigation strategies is lacking, especially for multi-turn LLM interactions. In this paper, we systematically investigate LLM vulnerabilities against prompt leakage for 10 closed- and open-source LLMs, across four domains. We design a unique threat model which leverages the LLM sycophancy effect and elevates the average attack success rate ("},"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":"2404.16251","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2024-04-24T23:39:58Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"1696394d4893ae5afbeafa36da87321f15bbf3f9d7ec19a573d5477d51adbafb","abstract_canon_sha256":"97189838dd644e70fcfa4620f3dd9e2970742b71934e36877ed79976aa0311b0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:49:33.541083Z","signature_b64":"2T56vKSaP6uX0oFZ6BC9K7XE/noEsM0ha/hNb190bwcZ5XPTNb0QgHLXWq4pX7Sa6Ri/3qF0Ni5BqKaoLm+fCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f853b7da7259434c2acdfa8e8560473cece8db365e99d9cb96d49acd5b3fe03d","last_reissued_at":"2026-07-05T08:49:33.540637Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:49:33.540637Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Prompt Leakage effect and defense strategies for multi-turn LLM interactions","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.CR","authors_text":"Alexander R. Fabbri, Ben Risher, Chien-Sheng Wu, Divyansh Agarwal, Philippe Laban, Shafiq Joty","submitted_at":"2024-04-24T23:39:58Z","abstract_excerpt":"Prompt leakage poses a compelling security and privacy threat in LLM applications. Leakage of system prompts may compromise intellectual property, and act as adversarial reconnaissance for an attacker. A systematic evaluation of prompt leakage threats and mitigation strategies is lacking, especially for multi-turn LLM interactions. In this paper, we systematically investigate LLM vulnerabilities against prompt leakage for 10 closed- and open-source LLMs, across four domains. We design a unique threat model which leverages the LLM sycophancy effect and elevates the average attack success rate ("},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.16251","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/2404.16251/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":"2404.16251","created_at":"2026-07-05T08:49:33.540701+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.16251v3","created_at":"2026-07-05T08:49:33.540701+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.16251","created_at":"2026-07-05T08:49:33.540701+00:00"},{"alias_kind":"pith_short_12","alias_value":"7BJ3PWTSLFBU","created_at":"2026-07-05T08:49:33.540701+00:00"},{"alias_kind":"pith_short_16","alias_value":"7BJ3PWTSLFBUYKWN","created_at":"2026-07-05T08:49:33.540701+00:00"},{"alias_kind":"pith_short_8","alias_value":"7BJ3PWTS","created_at":"2026-07-05T08:49:33.540701+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.27930","citing_title":"An LLM-Powered Semantic Alignment Framework for Journal Recommendation","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18231","citing_title":"AgenTEE: Confidential LLM Agent Execution on Edge Devices","ref_index":4,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7BJ3PWTSLFBUYKWN7KHIKYCHHT","json":"https://pith.science/pith/7BJ3PWTSLFBUYKWN7KHIKYCHHT.json","graph_json":"https://pith.science/api/pith-number/7BJ3PWTSLFBUYKWN7KHIKYCHHT/graph.json","events_json":"https://pith.science/api/pith-number/7BJ3PWTSLFBUYKWN7KHIKYCHHT/events.json","paper":"https://pith.science/paper/7BJ3PWTS"},"agent_actions":{"view_html":"https://pith.science/pith/7BJ3PWTSLFBUYKWN7KHIKYCHHT","download_json":"https://pith.science/pith/7BJ3PWTSLFBUYKWN7KHIKYCHHT.json","view_paper":"https://pith.science/paper/7BJ3PWTS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.16251&json=true","fetch_graph":"https://pith.science/api/pith-number/7BJ3PWTSLFBUYKWN7KHIKYCHHT/graph.json","fetch_events":"https://pith.science/api/pith-number/7BJ3PWTSLFBUYKWN7KHIKYCHHT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7BJ3PWTSLFBUYKWN7KHIKYCHHT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7BJ3PWTSLFBUYKWN7KHIKYCHHT/action/storage_attestation","attest_author":"https://pith.science/pith/7BJ3PWTSLFBUYKWN7KHIKYCHHT/action/author_attestation","sign_citation":"https://pith.science/pith/7BJ3PWTSLFBUYKWN7KHIKYCHHT/action/citation_signature","submit_replication":"https://pith.science/pith/7BJ3PWTSLFBUYKWN7KHIKYCHHT/action/replication_record"}},"created_at":"2026-07-05T08:49:33.540701+00:00","updated_at":"2026-07-05T08:49:33.540701+00:00"}