{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:CPWBRHATZL3TN6PA6BA2UNH2FD","merge_version":"pith-open-graph-merge-v1","event_count":3,"valid_event_count":3,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"8a981f444072b094ad1688b6e0bcbe4c6bfae8da37ba84d24ae62a67f8434b2a","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2026-05-09T16:16:41Z","title_canon_sha256":"99cafecba82d31d45b38b7753fc5ed4b54dc3c30a522023f1bd61277cf64408c"},"schema_version":"1.0","source":{"id":"2605.09033","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.09033","created_at":"2026-05-20T00:00:41Z"},{"alias_kind":"arxiv_version","alias_value":"2605.09033v3","created_at":"2026-05-20T00:00:41Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.09033","created_at":"2026-05-20T00:00:41Z"},{"alias_kind":"pith_short_12","alias_value":"CPWBRHATZL3T","created_at":"2026-05-20T00:00:41Z"},{"alias_kind":"pith_short_16","alias_value":"CPWBRHATZL3TN6PA","created_at":"2026-05-20T00:00:41Z"},{"alias_kind":"pith_short_8","alias_value":"CPWBRHAT","created_at":"2026-05-20T00:00:41Z"}],"graph_snapshots":[{"event_id":"sha256:b86018f7bb0265bf8fab78cf27e36e5d286aa9d4fade09845e8a5c5d4fcf69d4","target":"graph","created_at":"2026-05-20T00:00:41Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"SHADOWMERGE achieves 93.8% average attack success rate, improving the best baseline by 50.3 absolute points, while having negligible impact on unrelated benign tasks."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The graph-memory system will extract, merge into the target anchor neighborhood, and retrieve the poisoned relation for the victim query when it shares the same query-activated anchor and canonicalized relation channel as benign evidence via the AIR pipeline."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"ShadowMerge exploits relation-channel conflicts to poison graph-based agent memory, achieving 93.8% average attack success rate on Mem0 and real-world datasets while bypassing existing defenses."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"ShadowMerge poisons graph-based agent memory by injecting relations that share the same query-activated anchor and channel as legitimate evidence."}],"snapshot_sha256":"71a62ab75d1b668d9f05cfcf713abd301d691040a175eaee947fb2810fa46b9d"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":false,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T20:38:50.422746Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T13:31:19.305443Z","status":"completed","version":"1.0.0"},{"findings_count":1,"name":"doi_compliance","ran_at":"2026-05-19T10:32:52.510765Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.09033/integrity.json","findings":[{"audited_at":"2026-05-19T10:32:52.510765Z","detected_arxiv_id":null,"detected_doi":"10.1109/tkde.2022.3201246","detector":"doi_compliance","finding_type":"unresolvable_identifier","note":"Identifier '10.1109/tkde.2022.3201246' is syntactically valid but the DOI registry (doi.org) returned 404, and Crossref / OpenAlex / internal corpus also have no record. The cited work could not be located through any authoritative source.","ref_index":31,"severity":"critical","verdict_class":"cross_source"}],"snapshot_sha256":"730e1988477db5891f0e3a5bf459d60d0855b063d73ee968718398565890fcdd","summary":{"advisory":0,"by_detector":{"doi_compliance":{"advisory":0,"critical":1,"informational":0,"total":1}},"critical":1,"informational":0}},"paper":{"abstract_excerpt":"Graph-based agent memory is increasingly used in LLM agents to support structured long-term recall and multi-hop reasoning, but it also creates a new poisoning surface: an attacker can inject a crafted relation into graph memory so that it is later retrieved and influences agent behavior. Existing agent-memory poisoning attacks mainly target flat textual records and are ineffective in graph-based memory because malicious relations often fail to be extracted, merged into the target anchor neighborhood, or retrieved for the victim query.\n  We present SHADOWMERGE, a poisoning attack against graph","authors_text":"Lingyun Peng, Shuyu Li, Tiantian Ji, Xinran Liu, Yang Luo, Yong Liu, Zifeng Kang","cross_cats":["cs.AI"],"headline":"ShadowMerge poisons graph-based agent memory by injecting relations that share the same query-activated anchor and channel as legitimate evidence.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2026-05-09T16:16:41Z","title":"ShadowMerge: A Novel Poisoning Attack on Graph-Based Agent Memory via Relation-Channel Conflicts"},"references":{"count":57,"internal_anchors":14,"resolved_work":57,"sample":[{"cited_arxiv_id":"2210.03629","doi":"","is_internal_anchor":true,"ref_index":1,"title":"ReAct: Synergizing Reasoning and Acting in Language Models","work_id":"407a2351-25f1-497d-b611-f77d0292a8e6","year":2023},{"cited_arxiv_id":"2302.04761","doi":"","is_internal_anchor":true,"ref_index":2,"title":"Toolformer: Language Models Can Teach Themselves to Use Tools","work_id":"9bce40c8-cfd7-4983-80e0-c3bd4402322a","year":2023},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Re- flexion: Language agents with verbal reinforcement learning","work_id":"f1386d00-b17b-4a6f-ab98-78dc28287803","year":2023},{"cited_arxiv_id":"2305.16291","doi":"","is_internal_anchor":true,"ref_index":4,"title":"Voyager: An Open-Ended Embodied Agent with Large Language Models","work_id":"ffe0d207-86cf-4742-a100-e988ac8b9676","year":2023},{"cited_arxiv_id":"","doi":"10.1145/3586183.3606763","is_internal_anchor":false,"ref_index":5,"title":"Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology , year=","work_id":"b5bf85fe-4fb7-4966-b0b2-9ccf9d3b11b9","year":2023}],"snapshot_sha256":"06b7cefea1bead7d556bd25bde490e5efa4b46d5e106c2a890a44ce08c996140"},"source":{"id":"2605.09033","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-19T17:34:08.512884Z","id":"203f6772-45e7-42a7-8954-de7a118296b6","model_set":{"reader":"grok-4.3"},"one_line_summary":"ShadowMerge exploits relation-channel conflicts to poison graph-based agent memory, achieving 93.8% average attack success rate on Mem0 and real-world datasets while bypassing existing defenses.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"ShadowMerge poisons graph-based agent memory by injecting relations that share the same query-activated anchor and channel as legitimate evidence.","strongest_claim":"SHADOWMERGE achieves 93.8% average attack success rate, improving the best baseline by 50.3 absolute points, while having negligible impact on unrelated benign tasks.","weakest_assumption":"The graph-memory system will extract, merge into the target anchor neighborhood, and retrieve the poisoned relation for the victim query when it shares the same query-activated anchor and canonicalized relation channel as benign evidence via the AIR pipeline."}},"verdict_id":"203f6772-45e7-42a7-8954-de7a118296b6"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:026b472d0c6c48606a0b71554230b51bab99c1d0891febce998c74ae44f1f97b","target":"record","created_at":"2026-05-20T00:00:41Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"8a981f444072b094ad1688b6e0bcbe4c6bfae8da37ba84d24ae62a67f8434b2a","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2026-05-09T16:16:41Z","title_canon_sha256":"99cafecba82d31d45b38b7753fc5ed4b54dc3c30a522023f1bd61277cf64408c"},"schema_version":"1.0","source":{"id":"2605.09033","kind":"arxiv","version":3}},"canonical_sha256":"13ec189c13caf736f9e0f041aa34fa28e16fe313a39df369e6f1e165f678f80e","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"13ec189c13caf736f9e0f041aa34fa28e16fe313a39df369e6f1e165f678f80e","first_computed_at":"2026-05-20T00:00:41.867279Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:00:41.867279Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"/ARjyAy54JDYO0Obm9ZYtHwEn7VXUTWp1j0ha/rY5u+AnGKul6nzAoSoz7gEvXgrksYyZongBfmy0XlfBjLyAA==","signature_status":"signed_v1","signed_at":"2026-05-20T00:00:41.868065Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.09033","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:ffaf0bea5d58956f1b3bf94538483c8304a867b520ff8a83aa7aeb45f50de19f","sha256:026b472d0c6c48606a0b71554230b51bab99c1d0891febce998c74ae44f1f97b","sha256:b86018f7bb0265bf8fab78cf27e36e5d286aa9d4fade09845e8a5c5d4fcf69d4"],"state_sha256":"02d48137a1118ac1e06ab256959a9b4e1030eace47d5dfcbd816aab42795cfc7"}