The adversarial robustness of image quality assessment (IQA) models to adversarial attacks is emerging as a critical issue. Adversarial training has been widely used to improve the robustness of neural networks to adversarial attacks, but little in-depth research has examined adversarial training as a way to improve IQA model robustness. This study introduces an enhanced adversarial training approach tailored to IQA models; it adjusts the perceptual quality scores of adversarial images during training to enhance the correlation between an IQA model’s quality and the subjective quality scores. We also propose a new method for comparing IQA model robustness by measuring the Integral Robustness Score; this method evaluates the IQA model resistance to a set of adversarial perturbations with different magnitudes. We used our adversarial training approach to increase the robustness of five IQA models. Additionally, we tested the robustness of adversarially trained IQA models to 16 adversarial attacks and conducted an empirical probabilistic estimation of this feature.
1 January 2022 in Proceedings of the Institute for System Programming of the RAS
Our paper compares the accuracy of the vanilla ResNet-18 model with the accuracy of the Clipped BagNet-33 and BagNet-33 models with adversarial learning under different conditions. We performed experiments on images attacked by the adversarial sticker under conditions of image transformations. The adversarial sticker is a small region of the attacked image, inside which the pixel values can be changed indefinitely, and this can generate errors in the model prediction. The transformations of the attacked images in this paper simulate the distortions that appear in the physical world when a change in perspective, scale or lighting changes the image. Our experiments show that models from the BagNet family perform poorly on images in low quality. We also analyzed the effects of different types of transformations on the models' robustness to adversarial attacks and the tolerance of these attacks.
The adversarial robustness of image quality assessment (IQA) models to adversarial attacks is emerging as a critical issue. Adversarial training has been widely used to improve the robustness of neural networks to adversarial attacks, but little in-depth research has examined adversarial training as a way to improve IQA model robustness. This study introduces an enhanced adversarial training approach tailored to IQA models; it adjusts the perceptual quality scores of adversarial images during training to enhance the correlation between an IQA model’s quality and the subjective quality scores. We also propose a new method for comparing IQA model robustness by measuring the Integral Robustness Score; this method evaluates the IQA model resistance to a set of adversarial perturbations with different magnitudes. We used our adversarial training approach to increase the robustness of five IQA models. Additionally, we tested the robustness of adversarially trained IQA models to 16 adversarial attacks and conducted an empirical probabilistic estimation of this feature. The code is available at https://github.com/wianluna/metrics_at.
1 October 2024 in Journal of Mathematical Sciences
Modern large language models are huge systems with complex internal mechanisms implementing black-box response generation. Although aligned large language models have built-in defense mechanisms against attacks, recent studies demonstrate the vulnerability of large language models to attacks. In this study, we aim to expand the existing malicious datasets obtained from attacks so that similar vulnerabilities in large language models can be addressed in the future through the alignment procedure. In addition, we conduct experiments with modern large language models on our malicious dataset, which demonstrates the existing weaknesses in the models.
1 December 2025 in Programming and Computer Software
State-of-the-art (SOTA) large language models (LLMs) are huge systems with complex internal mechanisms that implement black-box response generation. Even though aligned LLMs have built-in defense mechanisms against attacks, recent studies demonstrate the vulnerability of LLMs to jailbreak attacks. In this study, we aim to extend the existing red-teaming datasets obtained from jailbreak attacks to address these LLM vulnerabilities in the future. In addition, we carry out some experiments with SOTA LLMs on our dataset to demonstrate the existing weaknesses in these models.