CAPTCHA recognition using machine learning algorithms with various techniques
Ádám Kovács, Tibor Gábor Tajti · Az Eszterházy Károly Tanárképző Főiskola tudományos közleményei. Tanulmányok a matematikai tudományok köréből/Az Eszterházy Károly Főiskola tudományos közleményei. Tanulmányok a matematikai tudományok köréből/Annales mathematicae et informaticae · 2023
In this paper, we present research results on the recognition of text-based CAPTCHA tests using advanced machine learning algorithms and techniques. Text-based CAPTCHAs serve as a crucial security measure to prevent automated access to various web services, but their effectiveness depends on their resistance to sophisticated recognition techniques. To this end, we focus on evaluating and enhancing the performance of recognition models using a Convolutional Neural Network (CNN) as the base model. We propose an integrated approach, which incorporates a systematic parameter optimization strategy using Grid Search Cross-Validation (Grid Search CV) and the Ensemble Voting Method to improve the performance of the recognition model. The use of Grid Search CV enables us to fine-tune the hyperparameters of the CNN model, leading to an optimal configuration. Further, we investigate the effectiveness of the Ensemble Voting Method to aggregate the predictions from multiple CNN models, each with a set of the optimal parameters obtained from the Grid Search CV. The methods’ performance was evaluated through multiple learning sessions, assessing their effectiveness in recognizing text-based CAPTCHAs under various scenarios.