FAWA: Fast Adversarial Watermark Attack on Optical Character Recognition (OCR) Systems
This work addresses the vulnerability of OCR systems to adversarial attacks for users who rely on OCR for critical applications, providing a more efficient and stealthy attack method.
This paper introduces FAWA, a Fast Adversarial Watermark Attack designed to generate natural-looking adversarial examples against sequence-based Optical Character Recognition (OCR) systems. The attack achieves a 100% success rate while using 60% fewer perturbations and 78% fewer iterations on average compared to existing methods, by disguising perturbations as watermarks.
Deep neural networks (DNNs) significantly improved the accuracy of optical character recognition (OCR) and inspired many important applications. Unfortunately, OCRs also inherit the vulnerabilities of DNNs under adversarial examples. Different from colorful vanilla images, text images usually have clear backgrounds. Adversarial examples generated by most existing adversarial attacks are unnatural and pollute the background severely. To address this issue, we propose the Fast Adversarial Watermark Attack (FAWA) against sequence-based OCR models in the white-box manner. By disguising the perturbations as watermarks, we can make the resulting adversarial images appear natural to human eyes and achieve a perfect attack success rate. FAWA works with either gradient-based or optimization-based perturbation generation. In both letter-level and word-level attacks, our experiments show that in addition to natural appearance, FAWA achieves a 100% attack success rate with 60% less perturbations and 78% fewer iterations on average. In addition, we further extend FAWA to support full-color watermarks, other languages, and even the OCR accuracy-enhancing mechanism.