{"id":"W4396243913","doi":"10.1145/3658644.3690194","title":"Evaluations of Machine Learning Privacy Defenses are Misleading","year":2024,"lang":"en","type":"preprint","venue":"","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":14,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Canadian Institute for Advanced Research; Universitas Brawijaya; Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung; National Science Foundation","keywords":"Internet privacy; Computer science; Computer security; Business; Psychology; Artificial intelligence","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","open_science","research_integrity"],"consensus_categories":[],"category_scores_codex":[0.001055,0.0003968505,0.0005691072,0.0004945995,0.0001851674,0.0003097046,0.001978413,0.0002719552,0.0001596618],"category_scores_gemma":[0.001604844,0.000372272,0.0002952557,0.0005349661,0.00007245567,0.0001692494,0.008806898,0.002352512,0.0001352705],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001401792,"about_ca_system_score_gemma":0.0003261082,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002313559,"about_ca_topic_score_gemma":0.00002699248,"domain_scores_codex":[0.9969341,0.0003754575,0.0006341211,0.0009682395,0.0007337941,0.0003543113],"domain_scores_gemma":[0.9974223,0.0005087522,0.0005808717,0.001126703,0.0002655653,0.00009580561],"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":[0.000006899723,0.0000610812,0.004552703,0.0009553437,0.0002425459,0.00004694933,0.004681659,0.9134611,0.000426567,0.04846783,0.0007326336,0.02636467],"study_design_scores_gemma":[0.0001541345,0.00003106863,0.000740117,0.0006926826,0.00009377691,0.00001367186,0.00007702255,0.9579657,0.0005886664,0.03805971,0.001174341,0.0004091427],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01369538,0.001581046,0.968701,0.002036872,0.001852119,0.0003899918,0.000008862832,0.001019777,0.01071493],"genre_scores_gemma":[0.8204701,0.00004249984,0.1758954,0.00005721409,0.0001808621,0.00002826197,0.00002139603,0.00004868112,0.00325554],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.8067747,"threshold_uncertainty_score":0.9999491,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03983493120052315,"score_gpt":0.3299973261049216,"score_spread":0.2901623949043984,"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."}}