ScalableCertifiedSegmentationviaRandomizedSmoothingMarcFischer1MaximilianBaader1MartinVechev1Abstract(a)Attackedimage(b)GroundtruthsegmentationWepresentanewcertificationmethodforimage(c)Attackedseg...
Improved,DeterministicSmoothingfor1CertifiedRobustnessAlexanderLevine1SoheilFeizi1Abstracthaviorofmachinelearningsystemsundersmallinputper-turbations(Szegedyetal.,2013;Goodfellowetal.,2014;Randomiz...
ScalableDifferentialPrivacywithCertifiedRobustnessinAdversarialLearningNhatHaiPhan1MyT.Thai2HanHu1RuomingJin3TongSun4DejingDou56Abstracttrainingdatatoconsistofbenignexamplesandadversar-ialexamplesc...
CertifiedDataRemovalfromMachineLearningModelsChuanGuo1TomGoldstein2AwniHannun2LaurensvanderMaaten2Abstractinferenceattacks(Yeometal.,2018;Carlinietal.,2019)areunsuccessfulondatathatwasremovedfromth...
CertifiedRobustnesstoLabel-FlippingAttacksviaRandomizedSmoothingElanRosenfeld1EzraWinston1PradeepRavikumar1J.ZicoKolter12Abstractuser-generateddatasets—e.g.emailspamfilters,productrecommendationen...
CertifiedAdversarialRobustnessviaRandomizedSmoothingJeremyCohen1ElanRosenfeld1J.ZicoKolter12AbstractpAWeshowhowtoturnanyclassifierthatclassifiesxpBwellunderGaussiannoiseintoanewclassifierthatiscert...
TowardsFastComputationofCertifiedRobustnessforReLUNetworksTsui-WeiWeng1HuanZhang2HonggeChen1ZhaoSong34Cho-JuiHsieh2DuaneBoning1InderjitS.Dhillon4LucaDaniel1Abstract1.IntroductionVerifyingtherobustn...