Adversary Detection in Online Machine Learning Systems

dc.contributorAalto-yliopistofi
dc.contributorAalto Universityen
dc.contributor.advisorMarchal, Samuel
dc.contributor.authorSzyller, Sebastian
dc.contributor.schoolPerustieteiden korkeakoulufi
dc.contributor.supervisorAsokan, N.
dc.date.accessioned2020-03-22T18:06:43Z
dc.date.available2020-03-22T18:06:43Z
dc.date.issued2020-03-16
dc.description.abstractMachine learning applications have become increasingly popular. At the same time, model training has become an expensive task in terms of computational power, amount of data, and human expertise. As a result, models now constitute intellectual property and business advantage to model owners and thus, their confidentiality must be preserved. Recently, it was shown that models can be stolen via model extraction attacks that do not require physical white-box access to the model but merely a black-box prediction API. Stolen model can be used to avoid paying for the service or even to undercut the offering of the legitimate model owner. Hence, it deprives the victim of the accumulated business advantage. In this thesis, we introduce two novel defense methods designed to detect distinct classes of model extraction attacks.en
dc.format.extent43
dc.identifier.urihttps://aaltodoc.aalto.fi/handle/123456789/43569
dc.identifier.urnURN:NBN:fi:aalto-202003222602
dc.language.isoenen
dc.programmeMaster’s Programme in Computer, Communication and Information Sciencesfi
dc.programme.majorMachine Learning, Data Science and Artificial Intelligencefi
dc.programme.mcodeSCI3044fi
dc.subject.keywordmachine learningen
dc.subject.keyworddeep learningen
dc.subject.keywordmodel extractionen
dc.subject.keywordIP protectionen
dc.titleAdversary Detection in Online Machine Learning Systemsen
dc.typeG2 Pro gradu, diplomityöfi
dc.type.ontasotMaster's thesisen
dc.type.ontasotDiplomityöfi
local.aalto.electroniconlyyes
local.aalto.openaccessno
Files