{"id":"W4383468997","doi":"10.48550/arxiv.2307.01610","title":"Overconfidence is a Dangerous Thing: Mitigating Membership Inference Attacks by Enforcing Less Confident Prediction","year":2023,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Overconfidence effect; Computer science; HAMP; Exploit; Inference; Machine learning; Training set; Artificial intelligence; Benchmark (surveying); Entropy (arrow of time); Data mining; Computer security","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.007522025,0.001082824,0.001198892,0.000684245,0.00120988,0.002170053,0.002445411,0.002127253,0.001114783],"category_scores_gemma":[0.03163365,0.0005549936,0.001006745,0.000709595,0.002545223,0.00519683,0.005977878,0.005351032,0.0005613712],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001170707,"about_ca_system_score_gemma":0.001430097,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0007481392,"about_ca_topic_score_gemma":0.0006955893,"domain_scores_codex":[0.990284,0.004766743,0.0003680096,0.001347618,0.002613801,0.0006198306],"domain_scores_gemma":[0.9741437,0.01191972,0.002360991,0.01004944,0.001017345,0.0005087082],"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":[0.001358094,0.000478356,0.0171521,0.000299157,0.0003964912,0.0008390045,0.001139118,0.4779837,0.03289923,0.1547477,0.01332306,0.2993839],"study_design_scores_gemma":[0.0000403651,0.0001401009,0.001125832,0.00004392648,0.00004195773,0.0003889172,0.00007854556,0.9358061,0.01676064,0.04250048,0.003033657,0.00003946725],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1282956,0.0009112611,0.8601881,0.00264871,0.0001408221,0.000109661,0.0002046363,0.002221311,0.005279881],"genre_scores_gemma":[0.9527663,0.0002172869,0.04498111,0.0005464194,0.000117039,0.00006028019,0.0001590597,0.00009426292,0.001058373],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.007522025,"threshold_uncertainty_score":0.0397808,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1314094578956852,"score_gpt":0.2432273903084682,"score_spread":0.111817932412783,"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."}}