{"id":"W4405468708","doi":"10.48550/arxiv.2412.10922","title":"Automatically Detecting Checked-In Secrets in Android Apps: How Far Are We?","year":2024,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Advanced Malware Detection Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Android (operating system); Computer science; Operating system","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0004338006,0.0004352058,0.0005160948,0.001094842,0.00007376864,0.0002233279,0.00166683,0.0004788744,0.000007935901],"category_scores_gemma":[0.0001374755,0.0005313447,0.0001914659,0.001675886,0.00008565687,0.0004650301,0.003491635,0.001685992,0.00003884146],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000739405,"about_ca_system_score_gemma":0.0001746984,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00008274587,"about_ca_topic_score_gemma":0.0006034242,"domain_scores_codex":[0.9972427,0.0001726095,0.0003166855,0.001599052,0.0001457682,0.0005231201],"domain_scores_gemma":[0.9981349,0.0001498186,0.0002891879,0.001185076,0.00009922696,0.0001417537],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0002497048,0.0009847664,0.01525041,0.006659363,0.0004376427,0.03878468,0.006500769,0.4523204,0.003975229,0.333704,0.0006746156,0.1404584],"study_design_scores_gemma":[0.0003478059,0.000059761,0.001126282,0.001113415,0.00001892408,0.00002851699,0.0002492261,0.7435992,0.003274118,0.2490955,0.0003629986,0.0007241663],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1578095,0.000270211,0.8373466,0.0004695818,0.0004543155,0.0005489477,0.000007829028,0.001941001,0.001152027],"genre_scores_gemma":[0.9838247,0.0003093015,0.01525365,0.00005763981,0.00003898242,0.000008530492,0.000001943376,0.00003760376,0.0004676066],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8260153,"threshold_uncertainty_score":0.9997138,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05929506377011211,"score_gpt":0.2060582314624557,"score_spread":0.1467631676923436,"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."}}