{"id":"W131327581","doi":"","title":"Security benchmarking using partial verification","year":2008,"lang":"en","type":"article","venue":"USENIX Security Symposium","topic":"Security and Verification in Computing","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Benchmarking; Computer science; Computer security; Order (exchange); Threat model; Partial order reduction; Model checking; Theoretical computer science; Business","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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0005824751,0.0003110142,0.0003107264,0.0002036079,0.0009964865,0.0002411915,0.001205136,0.0002141725,0.00003939603],"category_scores_gemma":[0.00006123156,0.0003599807,0.0001761822,0.0008231621,0.0001655009,0.001120312,0.0003803222,0.0004197938,0.00008931221],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001785419,"about_ca_system_score_gemma":0.0002201702,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001841211,"about_ca_topic_score_gemma":0.00000791011,"domain_scores_codex":[0.9969869,0.0002560275,0.0006042285,0.0009107244,0.0006251065,0.0006170087],"domain_scores_gemma":[0.9978861,0.0001453596,0.0003087041,0.00119709,0.000229052,0.0002337482],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0001542821,0.002600534,0.03499722,0.0004581164,0.0003352024,0.0006483675,0.1930657,0.007277778,0.07844581,0.6649531,0.01271886,0.004345003],"study_design_scores_gemma":[0.0006193643,0.00008814695,0.001120519,0.00005228998,0.00002188832,0.0006086908,0.00009986889,0.9399988,0.0235146,0.004602583,0.02850487,0.0007683748],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7454244,0.0002019724,0.246865,0.001108887,0.002795188,0.0003917512,0.000008818685,0.0007608237,0.002443118],"genre_scores_gemma":[0.9863042,0.00008584604,0.01238867,0.0003187268,0.0008266604,0.00001485546,0.00001741252,0.00002246767,0.00002116325],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.932721,"threshold_uncertainty_score":0.9998852,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0338729842272198,"score_gpt":0.2591525270121877,"score_spread":0.2252795427849679,"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."}}