Evaluating Adversarial Robustness on Document Image Classification

Timothée Fronteau, Arnaud Paran, Aymen Shabou

Abstract: Adversarial attacks and defenses have gained increasing interest on computer vision systems in recent years, but as of today, most investigations are limited to natural images. However, many artificial intelligence models actually handle documentary data, which is very different from real world images. Hence, in this work, we try to apply the adversarial attack philosophy on documentary data and to protect models against such attacks. Our methodology is to implement untargeted gradient-based, transfer-based and score-based attacks and evaluate the impact of defenses such as adversarial training, JPEG input compression and grey-scale input transformation on the robustness of ResNet50 and EfficientNetB0 model architectures. To the best of our knowledge, no such work has been conducted by the community in order to study the impact of these attacks on the document image classification task.