{"id":"W7123348570","doi":"10.1109/esem64174.2025.00021","title":"Secret Breach Detection in Source Code with Large Language Models","year":2025,"lang":"","type":"article","venue":"","topic":"Advanced Malware Detection Techniques","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"","keywords":"False positive paradox; Source code; Benchmark (surveying); Binary number; Scalability; Code (set theory); Binary code; Key (lock)","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.003611848,0.001755386,0.0007404305,0.002152556,0.0005297341,0.001640766,0.001586214,0.001070703,0.001013588],"category_scores_gemma":[0.01761068,0.0005223971,0.001428518,0.0008569412,0.0008064462,0.003064379,0.001762444,0.002090303,0.001400767],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001341279,"about_ca_system_score_gemma":0.001940463,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.009262135,"about_ca_topic_score_gemma":0.01275008,"domain_scores_codex":[0.9973768,0.0009806172,0.0002147722,0.000671069,0.0005657919,0.00019092],"domain_scores_gemma":[0.9880057,0.00754075,0.001130391,0.001719454,0.001363318,0.0002403166],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.001217689,0.0009150802,0.05035438,0.001042171,0.0004297628,0.001093606,0.0008286684,0.3768512,0.03741359,0.003169487,0.02788431,0.4988002],"study_design_scores_gemma":[0.00002561322,0.0001030136,0.002479956,0.00005069817,0.00004219094,0.0001776735,0.00007653079,0.9809998,0.01069408,0.002755727,0.002565818,0.0000288653],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6271392,0.003005999,0.302436,0.002363512,0.0002884555,0.0003925369,0.003356474,0.05731595,0.003701854],"genre_scores_gemma":[0.8658746,0.0005323797,0.1222482,0.0004805388,0.00007318459,0.0002165259,0.006697493,0.00116892,0.002708106],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.009262135,"threshold_uncertainty_score":0.0191015,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.006949916592829725,"score_gpt":0.2561849373312464,"score_spread":0.2492350207384166,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}