Proceedings of the 2020 IEEE International Conference on Software
Quality, Reliability, and Security
IEEE Computer Society Press, Los Alamitos, CA, pp. 406-413 (2020)
Zijie Li 1, 2 , Long Zhang 1, 2 , Jun Yan 3 , Jian Zhang 1, 2 , Zhenyu Zhang 4 , and T.H. Tse 5
[free download from QRS '20]
Deep neural networks (DNNs) have been widely used in classification tasks.
Studies have shown that DNNs
may be fooled by artificial examples known as adversaries.
A common technique for testing the robustness of a classification
is to apply perturbations (such as random noise) to existing
examples and try many of them iteratively, but it is very tedious
In this paper, we propose a technique to select adversaries more effectively.
We study the vulnerability of
examples by exploiting their class distinguishability.
In this way, we can evaluate the probability of generating adversaries from
each example, and prioritize all the examples accordingly.
We have conducted an empirical study using two DNN models on
four common datasets.
The results reveal that the vulnerability
of examples has a strong relationship with distinguishability.
The effectiveness of our technique is demonstrated through 79.67
to 99.68% improvements in the F-measure.
|EVERY VISITOR COUNTS:|