{"id":"W4391725330","doi":"10.14722/ndss.2024.23014","title":"Overconfidence is a Dangerous Thing: Mitigating Membership Inference Attacks by Enforcing Less Confident Prediction","year":2024,"lang":"en","type":"article","venue":"","topic":"Data Quality and Management","field":"Decision Sciences","cited_by":21,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Overconfidence effect; Inference; Computer security; Computer science; Artificial intelligence; Psychology; Social psychology","routes":{"ca_aff":true,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["scholarly_communication","insufficient_payload"],"consensus_categories":["insufficient_payload"],"category_scores_codex":[0.005142843,0.0002127424,0.0002676953,0.0002181508,0.0002447937,0.002531446,0.0009582082,0.0001054124,0.002895809],"category_scores_gemma":[0.001829877,0.0001651054,0.0001190132,0.0007189403,0.0001308343,0.001973519,0.0004491477,0.0003004288,0.0009035946],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00009430124,"about_ca_system_score_gemma":0.00009673211,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001670737,"about_ca_topic_score_gemma":0.0002808359,"domain_scores_codex":[0.9956563,0.0002201898,0.0008949094,0.0008816643,0.001963465,0.0003834397],"domain_scores_gemma":[0.9965861,0.00216337,0.0001603127,0.0007534281,0.000168957,0.0001677752],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"not_applicable","study_design_scores_codex":[0.00001456592,0.00004659457,0.001974327,0.000153867,0.0000919185,0.0000394718,0.01450435,0.0001644483,0.001849027,0.1704867,0.66954,0.1411347],"study_design_scores_gemma":[0.0004695886,0.000176794,0.0007812589,0.0008883218,0.0001024076,0.00002881428,0.05377673,0.0855592,0.01415903,0.06231232,0.7809095,0.0008360188],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1044768,0.00124634,0.6917713,0.008746517,0.003121643,0.0008651743,0.000470901,0.0009739318,0.1883274],"genre_scores_gemma":[0.973969,0.00006242809,0.0009204963,0.002938415,0.0001175686,0.00002901012,0.00003810622,0.00001397938,0.02191094],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8694922,"threshold_uncertainty_score":0.9998743,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2179136209609521,"score_gpt":0.4373605369772627,"score_spread":0.2194469160163106,"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."}}