{"id":"https://openalex.org/W4389713766","doi":"https://doi.org/10.48550/arxiv.2312.06674","title":"Guardrail Blind Spots on Turkish Prompt Injection: A Same-Set Miss-Rate Comparison of Four Guard Models","display_name":"Guardrail Blind Spots on Turkish Prompt Injection: A Same-Set Miss-Rate Comparison of Four Guard Models","publication_year":2023,"publication_date":"2023-12-07","ids":{"openalex":"https://openalex.org/W4389713766","doi":"https://doi.org/10.48550/arxiv.2312.06674"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2312.06674","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2312.06674","pdf_url":"https://arxiv.org/pdf/2312.06674","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2312.06674","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5071606441","display_name":"Hakan Inan","orcid":"https://orcid.org/0000-0003-3505-8838"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Inan, Hakan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5019445300","display_name":"Kartikeya Upasani","orcid":"https://orcid.org/0000-0001-8623-4243"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Upasani, Kartikeya","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5111524272","display_name":"Jianfeng Chi","orcid":"https://orcid.org/0009-0002-1138-5009"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chi, Jianfeng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5035660188","display_name":"Rashi Rungta","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Rungta, Rashi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5079492567","display_name":"Krithika Iyer","orcid":"https://orcid.org/0000-0003-2295-8618"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Iyer, Krithika","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5012370787","display_name":"Yuning Mao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mao, Yuning","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5093487556","display_name":"Michael Tontchev","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tontchev, Michael","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101778507","display_name":"Qing Hu","orcid":"https://orcid.org/0000-0001-8067-6569"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hu, Qing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5036265899","display_name":"Brian Fuller","orcid":"https://orcid.org/0000-0001-5865-2528"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fuller, Brian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5018125347","display_name":"Davide Testuggine","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Testuggine, Davide","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5054253075","display_name":"Madian Khabsa","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Khabsa, Madian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":52,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12262","display_name":"Hate Speech and Cyberbullying Detection","score":0.9706000089645386,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T12262","display_name":"Hate Speech and Cyberbullying Detection","score":0.9706000089645386,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T12151","display_name":"Interpreting and Communication in Healthcare","score":0.9320999979972839,"subfield":{"id":"https://openalex.org/subfields/3600","display_name":"General Health Professions"},"field":{"id":"https://openalex.org/fields/36","display_name":"Health Professions"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7105655670166016},{"id":"https://openalex.org/keywords/conversation","display_name":"Conversation","score":0.6023306846618652},{"id":"https://openalex.org/keywords/guard","display_name":"Guard (computer science)","score":0.5048578381538391},{"id":"https://openalex.org/keywords/moderation","display_name":"Moderation","score":0.4568851590156555},{"id":"https://openalex.org/keywords/taxonomy","display_name":"Taxonomy (biology)","score":0.42304155230522156},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.3864237666130066},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.32288432121276855},{"id":"https://openalex.org/keywords/machine-learning","display_name":"Machine learning","score":0.23883318901062012},{"id":"https://openalex.org/keywords/psychology","display_name":"Psychology","score":0.08237966895103455}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7105655670166016},{"id":"https://openalex.org/C2777200299","wikidata":"https://www.wikidata.org/wiki/Q52943","display_name":"Conversation","level":2,"score":0.6023306846618652},{"id":"https://openalex.org/C141141315","wikidata":"https://www.wikidata.org/wiki/Q2379942","display_name":"Guard (computer science)","level":2,"score":0.5048578381538391},{"id":"https://openalex.org/C93225998","wikidata":"https://www.wikidata.org/wiki/Q1941972","display_name":"Moderation","level":2,"score":0.4568851590156555},{"id":"https://openalex.org/C58642233","wikidata":"https://www.wikidata.org/wiki/Q8269924","display_name":"Taxonomy (biology)","level":2,"score":0.42304155230522156},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3864237666130066},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.32288432121276855},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.23883318901062012},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.08237966895103455},{"id":"https://openalex.org/C46312422","wikidata":"https://www.wikidata.org/wiki/Q11024","display_name":"Communication","level":1,"score":0.0},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.0},{"id":"https://openalex.org/C59822182","wikidata":"https://www.wikidata.org/wiki/Q441","display_name":"Botany","level":1,"score":0.0},{"id":"https://openalex.org/C86803240","wikidata":"https://www.wikidata.org/wiki/Q420","display_name":"Biology","level":0,"score":0.0}],"mesh":[],"locations_count":5,"locations":[{"id":"pmh:oai:arXiv.org:2312.06674","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2312.06674","pdf_url":"https://arxiv.org/pdf/2312.06674","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"pmh:doi:10.5281/zenodo.21908418","is_oa":true,"landing_page_url":"https://github.com/fevziegeyurtsevenler/guard-blindspots-tr","pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Report"},{"id":"doi:10.48550/arxiv.2312.06674","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2312.06674","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"},{"id":"doi:10.5281/zenodo.21908418","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.21908418","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Report"},{"id":"doi:10.5281/zenodo.21908419","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.21908419","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Report"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2312.06674","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2312.06674","pdf_url":"https://arxiv.org/pdf/2312.06674","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/16","score":0.7099999785423279,"display_name":"Peace, Justice and strong institutions"}],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4389713766.pdf","grobid_xml":"https://content.openalex.org/works/W4389713766.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W1968552888","https://openalex.org/W2374116601","https://openalex.org/W3093134843","https://openalex.org/W2772323916","https://openalex.org/W1511346092","https://openalex.org/W1527532029","https://openalex.org/W2378167147","https://openalex.org/W3210777354","https://openalex.org/W2281307425","https://openalex.org/W2464405057"],"abstract_inverted_index":{"Self-deposited":[0],"technical":[1],"report":[2],"\u2014":[3,192,301],"not":[4,68,154,206],"peer-reviewed.":[5],"On":[6],"a":[7,84,151,155,164,176],"single":[8],"fixed":[9],"set":[10],"of":[11,28,159],"248":[12,212],"Turkish":[13,81,188,200],"prompt-injection":[14,214],"attacks,":[15,30],"one":[16,71,158],"widely":[17],"used":[18],"open":[19,34],"guardrail":[20],"model,":[21],"jackhhao/jailbreak-classifier,":[22],"failed":[23],"to":[24,123],"flag":[25],"85.48%":[26],"(212/248)":[27,226],"the":[29,38,77,90,93,113,126,160],"while":[31],"two":[32],"other":[33],"guards":[35],"evaluated":[36],"on":[37,80,87,119,187,199],"exact":[39],"same":[40,127],"inputs":[41,89],"missed":[42,46,51,61],"almost":[43],"nothing:":[44],"fmops/distilbert-prompt-injection":[45,232],"0.0%":[47],"(0/248)":[48],"and":[49,95,140,167],"protectai/deberta-v3-base-prompt-injection-v2":[50,235],"1.61%":[52],"(4/248).":[53],"An":[54],"in-distribution":[55],"baseline":[56,170],"detector":[57],"we":[58],"trained":[59],"(altaysec-detector)":[60],"0.81%":[62],"(2/248).":[63],"The":[64,180],"headline":[65],"finding":[66],"is":[67,73,82,163,171,175,190],"that":[69,76],"any":[70],"guard":[72,97,198],"\"broken,\"":[74],"but":[75],"miss":[78,114],"rate":[79,115],"overwhelmingly":[83],"per-model":[85],"property:":[86],"identical":[88],"gap":[91],"between":[92],"worst":[94],"best":[96],"was":[98],"over":[99],"85":[100,271],"percentage":[101],"points.":[102],"We":[103,146],"further":[104],"probe":[105],"robustness":[106],"with":[107,132,136],"light":[108],"obfuscation.":[109],"Averaged":[110],"across":[111],"guards,":[112],"rises":[116],"from":[117],"21.26%":[118],"original-form":[120],"attacks":[121,128],"(n=107)":[122],"25.0%":[124],"when":[125],"are":[129],"broken":[130],"up":[131],"zero-width":[133],"spacing":[134],"(n=21),":[135],"casefolding":[137],"(22.92%,":[138],"n=60)":[139,143],"leetspeak":[141],"(21.25%,":[142],"in":[144],"between.":[145],"frame":[147],"these":[148],"numbers":[149],"as":[150],"disclosed-limit":[152],"measurement,":[153],"vendor":[156],"verdict:":[157],"four":[161],"models":[162],"jailbreak-specific":[165],"classifier":[166],"our":[168],"own":[169],"in-distribution,":[172],"so":[173],"neither":[174],"fair":[177],"head-to-head":[178],"winner.":[179],"practical":[181],"message":[182],"for":[183],"teams":[184],"deploying":[185],"LLMs":[186],"traffic":[189],"concrete":[191],"you":[193],"must":[194],"measure":[195],"your":[196],"specific":[197],"inputs,":[201],"because":[202],"English":[203],"benchmarks":[204],"do":[205],"transfer.":[207],"T\u00fcrk\u00e7e":[208,213,309,314],"\u00f6zet:":[209],"Ayn\u0131":[210],"sabit":[211],"sald\u0131r\u0131s\u0131":[215],"\u00fczerinde,":[216],"yayg\u0131n":[217],"kullan\u0131lan":[218],"a\u00e7\u0131k":[219],"bir":[220,250,305],"koruma":[221,306],"modeli":[222],"jackhhao/jailbreak-classifier":[223],"sald\u0131r\u0131lar\u0131n":[224],"%85,48'ini":[225],"yakalayamad\u0131;":[227],"oysa":[228],"ayn\u0131":[229,262],"girdiler":[230],"\u00fczerinde":[231],"%0,0":[233],"(0/248),":[234],"ise":[236,295],"%1,61":[237],"(4/248)":[238],"ka\u00e7\u0131rd\u0131.":[239,246],"E\u011fitti\u011fimiz":[240],"taban":[241,293],"\u00e7izgisi":[242],"altaysec-detector":[243],"%0,81":[244],"(2/248)":[245],"Ana":[247],"bulgu,":[248],"herhangi":[249],"koruman\u0131n":[251],"\"bozuk\"":[252],"olmas\u0131":[253],"de\u011fil,":[254],"ka\u00e7\u0131rma":[255,278],"oran\u0131n\u0131n":[256],"b\u00fcy\u00fck":[257],"\u00f6l\u00e7\u00fcde":[258],"modele":[259],"\u00f6zg\u00fc":[260],"olmas\u0131d\u0131r:":[261],"girdilerde":[263],"en":[264,267],"k\u00f6t\u00fc":[265],"ile":[266,276],"iyi":[268,303,310],"aras\u0131ndaki":[269],"fark":[270],"puan\u0131":[272],"a\u015ft\u0131.":[273],"Hafif":[274],"gizleme":[275],"ortalama":[277],"orijinalde":[279],"%21,26'dan":[280],"s\u0131f\u0131r-geni\u015flikli":[281],"bo\u015flukta":[282],"%25,0'a":[283],"\u00e7\u0131k\u0131yor.":[284],"K\u0131s\u0131tlar":[285],"a\u00e7\u0131k\u00e7a":[286],"belirtildi:":[287],"modellerden":[288],"biri":[289],"jailbreak'e":[290],"\u00f6zg\u00fc,":[291],"kendi":[292,312],"\u00e7izgimiz":[294],"da\u011f\u0131l\u0131m":[296],"i\u00e7idir.":[297],"Pratik":[298],"ders":[299],"nettir":[300],"\u0130ngilizce":[302],"olan":[304],"otomatik":[307],"olarak":[308],"de\u011fildir;":[311],"koruman\u0131z\u0131":[313],"girdilerle":[315],"\u00f6l\u00e7\u00fcn.":[316],"Open":[317],"data":[318],"&":[319,323],"code:":[320],"GitHub":[321],"(code":[322],"data,":[324],"open)":[325],"\u00b7":[326,330],"Hugging":[327],"Face":[328],"dataset":[329],"companion":[331],"article:":[332],"altaysec.com.tr.":[333],"Author:":[334],"Fevzi":[335],"Ege":[336],"Yurtsevenler,":[337],"AltaySec":[338],"(T\u00fcrkiye).":[339],"License:":[340],"CC":[341],"BY":[342],"4.0.":[343]},"counts_by_year":[{"year":2026,"cited_by_count":10},{"year":2025,"cited_by_count":28},{"year":2024,"cited_by_count":14}],"updated_date":"2026-08-22T07:34:49.880490","created_date":"2025-10-10T00:00:00"}