Variational Color Shift and Auto‐Encoder based on Large Separable Kernel Attention for Enhanced Text CAPTCHA Vulnerability Assessment

Text CAPTCHAs are crucial security measures employed on global websites to deter unauthorized intrusions. The presence of anti-attack features incorporated into text CAPTCHAs limits the effectiveness of breaking them, despite CAPTCHA recognition being an effective method for assessing their security. This study introduces a novel color augmentation technique called Variational Color Shift (VCS) to boost the recognition accuracy of different networks. VCS generates a color shift range of every input image and then resamples the image within that range to generate a new image, thus expanding the number of samples of the original dataset to improve training effectiveness. In contrast to Random Color Shift (RCS) which treats the color offsets as hyperparameters, VCS uses the estimated offsets to reparametrize the points sampling from the unit uniform distribution to reconstruct new image pixels, which makes the offsets learnable. To reduce the computational effort of VCS, we also propose two variants of VCS: Sim-VCS and Dilated-VCS. In addition, to solve the overfitting problem caused by many disturbances in CAPTCHAs, we propose an Auto Encoder (AE) based on a Large Separable Kernel Attention (AE-LSKA) to replace the convolutional module in the text CAPTCHA recognizer. This new module employs an AE) to compress the interference while expanding the receptive field using Large Separable Kernel Attention (LSKA), which reduces the impact of local interference on the model training and improves the overall perception of characters. The experimental results show that the recognition rate of the model after integrating the AE-LSKA module is improved by at least 15 percentage points on both M-CAPTCHA and P-CAPTCHA datasets. In addition, experimental results demonstrate that color augmentation using VCS is more effective in enhancing recognition.