{"id":"W4414982617","doi":"10.1016/j.media.2025.103807","title":"A novel gradient inversion attack framework to investigate privacy vulnerabilities during retinal image-based federated learning","year":2025,"lang":"en","type":"article","venue":"Medical Image Analysis","topic":"Privacy-Preserving Technologies in Data","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"Alberta Children's Hospital; University of Calgary","funders":"Canada Research Chairs","keywords":"Generalizability theory; Convolutional neural network; Deep learning; Data set; Differential privacy; Vulnerability (computing); Artificial neural network; Inversion (geology); Information privacy; Training set","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":["metaresearch","metaepi_narrow","open_science"],"consensus_categories":["open_science"],"category_scores_codex":[0.001485461,0.0003520695,0.0006235306,0.001264417,0.0006480376,0.0006790253,0.01405515,0.000305316,0.0002161439],"category_scores_gemma":[0.1481127,0.0003286204,0.000285466,0.007015378,0.0005316128,0.0006346804,0.03650961,0.001380013,0.00008028054],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003123205,"about_ca_system_score_gemma":0.000308737,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0003045231,"about_ca_topic_score_gemma":0.00004580129,"domain_scores_codex":[0.9957452,0.0003490131,0.0006823085,0.001252481,0.001215946,0.0007550933],"domain_scores_gemma":[0.9931357,0.0008284717,0.0001892131,0.00506522,0.00031499,0.0004664215],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0006247486,0.003477174,0.1839905,0.002556111,0.009630453,0.00340006,0.004773524,0.01240524,0.1601987,0.00618808,0.5307149,0.08204056],"study_design_scores_gemma":[0.0005910911,0.00007383372,0.004677191,0.0004243868,0.0002614307,0.000004539066,0.0001315936,0.9619012,0.02229972,0.007928271,0.001252641,0.0004540364],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.2276884,0.00004126662,0.6945763,0.0764377,0.0001135406,0.000138887,0.000005257645,0.0008199709,0.0001787162],"genre_scores_gemma":[0.4283445,0.00001695465,0.5692935,0.002027075,0.00003657786,0.00004555165,0.00003642559,0.00001515826,0.0001843176],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.949496,"threshold_uncertainty_score":0.9999166,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02079675078901683,"score_gpt":0.2987669005006848,"score_spread":0.277970149711668,"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."}}