{"id":"W4293093392","doi":"10.1109/pst55820.2022.9851981","title":"Careful What You Wish For: on the Extraction of Adversarially Trained Models","year":2022,"lang":"en","type":"preprint","venue":"","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"Polytechnique Montréal","funders":"Natural Sciences and Engineering Research Council of Canada; Synopsys","keywords":"Computer science; Adversarial system; Robustness (evolution); Artificial intelligence; Machine learning; Leverage (statistics); Adversarial machine learning; Adversary; Attack model; Computer security","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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.001447215,0.0003549405,0.0004433884,0.0001882937,0.0003177297,0.0003608585,0.002835409,0.0002451878,0.0002101749],"category_scores_gemma":[0.0002919485,0.0002796024,0.0003854848,0.0002399689,0.00006655667,0.0009616946,0.002174889,0.001266963,0.000002846843],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002537694,"about_ca_system_score_gemma":0.0003838132,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001520771,"about_ca_topic_score_gemma":0.00001852506,"domain_scores_codex":[0.9970813,0.0003617081,0.0005294787,0.0008961805,0.0007977977,0.0003335078],"domain_scores_gemma":[0.9965735,0.0009109883,0.0006122969,0.001647103,0.0001951938,0.00006088252],"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":[0.00004987143,0.00004409425,0.000002704018,0.00004382203,0.0000645172,0.000003966011,0.001692035,0.7741405,0.00004784349,0.2104704,0.0005776994,0.01286254],"study_design_scores_gemma":[0.0004060282,0.0001344684,0.00002362771,0.00008277342,0.00004013799,0.000004405451,0.0009545198,0.9072188,0.0001746863,0.08737722,0.003253505,0.0003299019],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.001646413,0.00008230376,0.9763542,0.005625879,0.004725104,0.001093125,0.00001882104,0.0002464658,0.0102077],"genre_scores_gemma":[0.8726719,0.0001293181,0.1227894,0.00114791,0.0005492287,0.000470823,0.00008577155,0.00007708945,0.00207848],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.8710255,"threshold_uncertainty_score":0.9999656,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05422428654648595,"score_gpt":0.3085886203618194,"score_spread":0.2543643338153335,"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."}}