学术讲座通知（4月22日）

讲座主题：Vulnerability,MisuseandProtectionofDeepNetworks
主讲人：Dr.AdamsWai-KinKong
讲座时间：2024年4月22日（周一）下午15:00-16:30
讲座地点：（略）
主讲人简介：Dr.AdamsWai-（略）,Canada.Currently,heisanassociateprofessorattheNanyangTechnologicalUniversity,（略）,oneofhispaperspublishedinCVPRwasselectedfororalpresentation.HisPhDstudentsreceivedBestStudentPaperAwardsinTheIEEEFifthInternationalConferenceonBiometrics:Theory,ApplicationsandSystems(BTAS),2012andIEEEInternationalConferenceonBioinformaticsandBioengineering,2013andotherawardssuchasGoogleAnitaBorgMemorialScholarship.HeservedasanassociateeditorforIEEETransactionsonInformationForensicsandSecurity,areachairofAIandCVconferences,includingEECVandIJCAIandexpertconsultantforcross-borderlegalcases.HeislistedinStanfordUniversity'sTop2%Scientists’Study.In2022,（略）,deeplearningandtheirapplicationsoncomputationalsimulation,healthcare,andbiometrics.
报告简介：Inrecentyears,deepnetworkshavedemonstratedtheirpowerfulcapabilities,oftenoutperforminghumanexperts,suchasAlphaGO.Manyhavebeendeployedtorealapplications,creatinggreatbusinessvalue.Becauseoftheirgreatimpacttohumanbeingsandoursociety,manyhaveconcernedabouttheirpotentialrisks.Inthistalk,thespeakerwillfirstgivesomebackgroundinformationaboutadversarialattacks,includingblack-boxandwhite-boxattacksandattributionmethods,includingintegratedgradients.Then,hewilldiscusshowtomisleadobjectdetectorsandperformastrongtransferableattack.Then,hewillfurtherdiscussissuesaboutmisusingimageinpaintingnetworksandimage-basedgenerativemodels.Takingvisiblewatermarkasanexample,hewillpresentamethodtoprotectwatermarksagainstthemisuseofinpaintingnetworks.Furthermore,hewilldemonstratehowtoremovethetoxicconceptsindiffusionnetworks.Finally,hewillbrieflymentionsomeoftheoreticalworksaboutattributionprotection.
欢迎广大师生积极参加！