{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:ATURPM3JGDKTQR5LQ53S7KAYIN","short_pith_number":"pith:ATURPM3J","schema_version":"1.0","canonical_sha256":"04e917b36930d53847ab87772fa818436c6f47df77756faf21066bf94114289c","source":{"kind":"arxiv","id":"2203.10346","version":1},"attestation_state":"computed","paper":{"title":"Perturbations in the Wild: Leveraging Human-Written Text Perturbations for Realistic Adversarial Attack and Defense","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.CR"],"primary_cat":"cs.LG","authors_text":"Dongwon Lee, Jooyoung Lee, Kevin Yen, Thai Le, Yifan Hu","submitted_at":"2022-03-19T16:00:01Z","abstract_excerpt":"We proposes a novel algorithm, ANTHRO, that inductively extracts over 600K human-written text perturbations in the wild and leverages them for realistic adversarial attack. Unlike existing character-based attacks which often deductively hypothesize a set of manipulation strategies, our work is grounded on actual observations from real-world texts. We find that adversarial texts generated by ANTHRO achieve the best trade-off between (1) attack success rate, (2) semantic preservation of the original text, and (3) stealthiness--i.e. indistinguishable from human writings hence harder to be flagged"},"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":"2203.10346","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-03-19T16:00:01Z","cross_cats_sorted":["cs.CL","cs.CR"],"title_canon_sha256":"c13b50acaeee363c6a1a62bce3af8cf5532f8ce55317d0fef006e103547c40fd","abstract_canon_sha256":"2a13614ead43dfff43e3e19327d02af51661d825c6f2212367d9ff804afffa74"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:06:51.995207Z","signature_b64":"YObxnLjzhGqy9MEcQcrTeFGdVN4WAGxhnPad5LcOAA9qx7GjK7jxN1q33EKoR3W1retEqww3ACwGXVpKfFCoDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"04e917b36930d53847ab87772fa818436c6f47df77756faf21066bf94114289c","last_reissued_at":"2026-07-05T04:06:51.994809Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:06:51.994809Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Perturbations in the Wild: Leveraging Human-Written Text Perturbations for Realistic Adversarial Attack and Defense","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.CR"],"primary_cat":"cs.LG","authors_text":"Dongwon Lee, Jooyoung Lee, Kevin Yen, Thai Le, Yifan Hu","submitted_at":"2022-03-19T16:00:01Z","abstract_excerpt":"We proposes a novel algorithm, ANTHRO, that inductively extracts over 600K human-written text perturbations in the wild and leverages them for realistic adversarial attack. Unlike existing character-based attacks which often deductively hypothesize a set of manipulation strategies, our work is grounded on actual observations from real-world texts. We find that adversarial texts generated by ANTHRO achieve the best trade-off between (1) attack success rate, (2) semantic preservation of the original text, and (3) stealthiness--i.e. indistinguishable from human writings hence harder to be flagged"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2203.10346","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/2203.10346/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":"2203.10346","created_at":"2026-07-05T04:06:51.994864+00:00"},{"alias_kind":"arxiv_version","alias_value":"2203.10346v1","created_at":"2026-07-05T04:06:51.994864+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2203.10346","created_at":"2026-07-05T04:06:51.994864+00:00"},{"alias_kind":"pith_short_12","alias_value":"ATURPM3JGDKT","created_at":"2026-07-05T04:06:51.994864+00:00"},{"alias_kind":"pith_short_16","alias_value":"ATURPM3JGDKTQR5L","created_at":"2026-07-05T04:06:51.994864+00:00"},{"alias_kind":"pith_short_8","alias_value":"ATURPM3J","created_at":"2026-07-05T04:06:51.994864+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2508.18306","citing_title":"SALMAN: Stability Analysis of Language Models Through the Maps Between Graph-based Manifolds","ref_index":50,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ATURPM3JGDKTQR5LQ53S7KAYIN","json":"https://pith.science/pith/ATURPM3JGDKTQR5LQ53S7KAYIN.json","graph_json":"https://pith.science/api/pith-number/ATURPM3JGDKTQR5LQ53S7KAYIN/graph.json","events_json":"https://pith.science/api/pith-number/ATURPM3JGDKTQR5LQ53S7KAYIN/events.json","paper":"https://pith.science/paper/ATURPM3J"},"agent_actions":{"view_html":"https://pith.science/pith/ATURPM3JGDKTQR5LQ53S7KAYIN","download_json":"https://pith.science/pith/ATURPM3JGDKTQR5LQ53S7KAYIN.json","view_paper":"https://pith.science/paper/ATURPM3J","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2203.10346&json=true","fetch_graph":"https://pith.science/api/pith-number/ATURPM3JGDKTQR5LQ53S7KAYIN/graph.json","fetch_events":"https://pith.science/api/pith-number/ATURPM3JGDKTQR5LQ53S7KAYIN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ATURPM3JGDKTQR5LQ53S7KAYIN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ATURPM3JGDKTQR5LQ53S7KAYIN/action/storage_attestation","attest_author":"https://pith.science/pith/ATURPM3JGDKTQR5LQ53S7KAYIN/action/author_attestation","sign_citation":"https://pith.science/pith/ATURPM3JGDKTQR5LQ53S7KAYIN/action/citation_signature","submit_replication":"https://pith.science/pith/ATURPM3JGDKTQR5LQ53S7KAYIN/action/replication_record"}},"created_at":"2026-07-05T04:06:51.994864+00:00","updated_at":"2026-07-05T04:06:51.994864+00:00"}