{"id":"W4401990434","doi":"10.1109/dsn-w60302.2024.00021","title":"Unlearning Backdoor Attacks Through Gradient-Based Model Pruning","year":2024,"lang":"en","type":"article","venue":"","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Queensland University of Technology; Canadian Institute for Advanced Research","keywords":"Backdoor; Computer science; Pruning; Artificial intelligence; 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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0004760319,0.0002482487,0.0002045184,0.0001578301,0.0002867991,0.000570529,0.001014719,0.00009922763,0.0001108833],"category_scores_gemma":[0.0001278919,0.0002203363,0.0001343473,0.0007855695,0.00006143447,0.001301223,0.0004133509,0.0006310811,0.0003020874],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001124991,"about_ca_system_score_gemma":0.0002152536,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0000405242,"about_ca_topic_score_gemma":0.000008566323,"domain_scores_codex":[0.9979244,0.0001092985,0.0002827614,0.000746663,0.0004255805,0.0005112881],"domain_scores_gemma":[0.9989079,0.0002807799,0.00005801698,0.0005989311,0.00005551755,0.00009891811],"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":[0.000002326161,0.00001189304,0.0001315047,0.00003491811,0.00001495315,0.00004118843,0.0008493545,0.7549901,0.0001529097,0.2353237,0.0009941572,0.007453004],"study_design_scores_gemma":[0.0001928288,0.00003532577,0.00002554136,0.00009951196,0.00001068161,0.00000998657,0.00004825552,0.9851261,0.0002780198,0.005032079,0.008854063,0.0002875495],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.001969025,0.0002056259,0.9646408,0.00218514,0.0007728466,0.0001283167,6.427978e-7,0.001750119,0.0283475],"genre_scores_gemma":[0.5742472,0.000003300501,0.4222174,0.0005312884,0.00009991189,0.00001047043,0.000003683652,0.0000313895,0.00285541],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.5722781,"threshold_uncertainty_score":0.8985053,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02581618714497178,"score_gpt":0.2994003539497049,"score_spread":0.2735841668047332,"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."}}