{"id":"W4408897262","doi":"10.1109/icmlc63072.2024.10935257","title":"The Bifurcation Method: White-Box Observation Perturbation Attacks on Reinforcement Learning Agents on a Cyber Physical System","year":2024,"lang":"en","type":"article","venue":"","topic":"Advanced Research in Systems and Signal Processing","field":"Engineering","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"Polytechnique Montréal; Royal Military College of Canada","funders":"","keywords":"White box; Reinforcement learning; Bifurcation; Cyber-physical system; Computer science; Perturbation (astronomy); Reinforcement; Artificial intelligence; Control theory (sociology); Physics; Engineering; Machine learning; Structural engineering; Nonlinear system; Operating system; Quantum mechanics","routes":{"ca_aff":true,"ca_fund":false,"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":[],"consensus_categories":[],"category_scores_codex":[0.0005304033,0.0001480128,0.0001156655,0.00008215724,0.0003002605,0.0002654226,0.00009822073,0.00005416786,0.00001329315],"category_scores_gemma":[0.0000621838,0.00009273142,0.00006043115,0.0002739697,0.00001210279,0.0002417005,0.00001640815,0.000334018,0.0001459694],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003678716,"about_ca_system_score_gemma":0.00001778175,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000005338388,"about_ca_topic_score_gemma":7.177996e-7,"domain_scores_codex":[0.998777,0.00007787481,0.0002333759,0.0002056692,0.0004707264,0.0002354064],"domain_scores_gemma":[0.9993873,0.0003047336,0.00003064809,0.0001540671,0.00006885033,0.00005443827],"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":[0.00001298095,0.000005692171,0.00001583015,0.0003033559,0.00003409946,0.000002561106,0.0006858426,0.9397867,0.002446603,0.02027377,0.0009462725,0.03548629],"study_design_scores_gemma":[0.00008044736,0.0000826672,0.00009696539,0.0004397182,0.000006728827,0.000001625407,0.0004893375,0.9753643,0.003065947,0.0001480263,0.02011047,0.0001137649],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02312984,0.0002072345,0.9030703,0.0003920186,0.0007967198,0.0005414434,9.668346e-7,0.001013659,0.07084782],"genre_scores_gemma":[0.9868093,0.00002015626,0.0005689262,0.00002504332,0.0003452116,0.00009475966,0.00001695515,0.00003773913,0.01208188],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9636795,"threshold_uncertainty_score":0.3781478,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03741217608528022,"score_gpt":0.3286946572046233,"score_spread":0.2912824811193431,"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."}}