Defending against Poisoning Attacks in Online Learning Settings

File Description SizeFormat 
Defending against Poisoning Attacks in Online Learning Settings.pdfAccepted version196.88 kBAdobe PDFView/Open
Title: Defending against Poisoning Attacks in Online Learning Settings
Authors: Collinge, G
Lupu, E
Munoz Gonzalez, L
Item Type: Conference Paper
Abstract: Machine learning systems are vulnerable to data poisoning, a coordinated attack where a fraction of the training dataset is manipulated by an attacker to subvert learning. In this paper we first formulate an optimal attack strategy against online learning classifiers to assess worst-case scenarios. We also propose two defence mechanisms to mitigate the effect of online poisoning attacks by analysing the impact of the data points in the classifier and by means of an adaptive combination of machine learning classifiers with different learning rates. Our experimental evaluation supports the usefulness of our proposed defences to mitigate the effect of poisoning attacks in online learning settings.
Issue Date: 28-Mar-2019
Date of Acceptance: 24-Jan-2019
URI: http://hdl.handle.net/10044/1/70348
ISBN: 9782875870650
Publisher: ESANN
Journal / Book Title: Proceedings of the 27th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning
Copyright Statement: © 2019 ESANN. All rights reserved.
Conference Name: European Symposium on Artificial Neural Networks
Publication Status: Published
Start Date: 2019-04-24
Finish Date: 2019-04-26
Conference Place: Bruges, Belgium
Appears in Collections:Faculty of Engineering
Computing



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

Creative Commonsx