07795200.pdf (9.81 MB)
Improving the robustness of neural networks using K-support norm based adversarial training
Version 2 2023-06-12, 08:36
Version 1 2023-06-09, 04:36
journal contribution
posted on 2023-06-12, 08:36 authored by Sheikh Akhtar, Saad Rehman, Mahmood Akthar, Muazzam Kahn, Farhan Riaz, Qaiser Chaudry, Rupert YoungRupert YoungIt is of significant importance for any classification and recognition system, which claims near or better than human performance to be immune to small perturbations in the dataset. Researchers found out that neural networks are not very robust to small perturbations and can easily be fooled to persistently misclassify by adding a particular class of noise in the test data. This, so-called adversarial noise severely deteriorates the performance of neural networks, which otherwise perform really well on unperturbed dataset. It has been recently proposed that neural networks can be made robust against adversarial noise by training them using the data corrupted with adversarial noise itself. Following this approach, in this paper, we propose a new mechanism to generate a powerful adversarial noise model based on K-support norm to train neural networks. We tested our approach on two benchmark datasets, namely the MNIST and STL-10, using muti-layer perceptron and convolutional neural networks. Experimental results demonstrate that neural networks trained with the proposed technique show significant improvement in robustness as compared to state-of-the-art techniques.
History
Publication status
- Published
File Version
- Published version
Journal
IEEE AccessISSN
2169-3536Publisher
IEEEExternal DOI
Issue
2016Volume
4Page range
9501-9511Department affiliated with
- Engineering and Design Publications
Research groups affiliated with
- Industrial Informatics and Signal Processing Research Group Publications
Full text available
- Yes
Peer reviewed?
- Yes
Legacy Posted Date
2017-01-09First Open Access (FOA) Date
2017-01-09First Compliant Deposit (FCD) Date
2017-01-08Usage metrics
Categories
No categories selectedLicence
Exports
RefWorks
BibTeX
Ref. manager
Endnote
DataCite
NLM
DC