Harnessing the Vulnerability of Latent Layers in Adversarially Trained Models

Sinha, Abhishek and Balasubramanian, Vineeth N and Machiraju, Harshitha et. al. (2019) Harnessing the Vulnerability of Latent Layers in Adversarially Trained Models. arXiv.


Download (661kB) | Preview


Neural networks are vulnerable to adversarial attacks - small visually imperceptible crafted noise which when added to the input drastically changes the output. The most effective method of defending against these adversarial attacks is to use the methodology of adversarial training. We analyze the adversarially trained robust models to study their vulnerability against adversarial attacks at the level of the latent layers. Our analysis reveals that contrary to the input layer which is robust to adversarial attack, the latent layer of these robust models are highly susceptible to adversarial perturbations of small magnitude. Leveraging this information, we introduce a new technique Latent Adversarial Training (LAT) which comprises of fine-tuning the adversarially trained models to ensure the robustness at the feature layers. We also propose Latent Attack (LA), a novel algorithm for construction of adversarial examples. LAT results in minor improvement in test accuracy and leads to a state-of-the-art adversarial accuracy against the universal first-order adversarial PGD attack which is shown for the MNIST, CIFAR-10, CIFAR-100 datasets.

[error in script]
IITH Creators:
IITH CreatorsORCiD
Balasubramanian, Vineeth NUNSPECIFIED
Item Type: Article
Subjects: Computer science
Divisions: Department of Computer Science & Engineering
Depositing User: Team Library
Date Deposited: 28 May 2019 07:58
Last Modified: 28 May 2019 07:58
URI: http://raiith.iith.ac.in/id/eprint/5345
Publisher URL:
Related URLs:

Actions (login required)

View Item View Item
Statistics for RAIITH ePrint 5345 Statistics for this ePrint Item