Gradient-Based Enhancement Attacks in Biomedical Machine Learning
Rosenblatt M, Dadashkarimi J, Scheinost D. Gradient-Based Enhancement Attacks in Biomedical Machine Learning. Lecture Notes In Computer Science 2023, 14242: 301-312. DOI: 10.1007/978-3-031-45249-9_29.Peer-Reviewed Original ResearchBiomedical machinePrediction performanceEnhancement frameworkData provenance trackingPrevalence of machineEnhanced dataHigh feature similaritySimple neural networkProvenance trackingAdversarial attacksNeural networkData manipulationFeature similarityMedical imagingMachineEnhanced datasetOriginal datasetClassifierAttacksDatasetTrustworthinessModel performancePrevious workFrameworkPerformanceConnectome-based machine learning models are vulnerable to subtle data manipulations
Rosenblatt M, Rodriguez R, Westwater M, Dai W, Horien C, Greene A, Constable R, Noble S, Scheinost D. Connectome-based machine learning models are vulnerable to subtle data manipulations. Patterns 2023, 4: 100756. PMID: 37521052, PMCID: PMC10382940, DOI: 10.1016/j.patter.2023.100756.Peer-Reviewed Original ResearchData manipulationNoise attacksPrediction performanceMachine learning modelsManipulated dataLearning modelHigh trustworthinessConnectome dataTrustworthinessAttacksModel performancePredictive modelDownstream analysisPerformanceAcademic researchMachineRobustnessModelConnectomeConnectome-based modelsFunctional connectomeManipulation