Research on Identification of Cryptographic Algorithms Based on BERT Model and BPNN

With the widespread adoption of Internet technology and advancements in computer application technology, information security issues in large-scale information networks, such as power networks, have become increasingly prevalent. To address the challenges posed by extensive user privacy data introduced by IoT terminals in power networks, this study focuses on the effective identification of passwords for power network information system terminals. By integrating the random forest algorithm with two machine learning algorithms, a novel feature recognition model for grouped passwords in power information systems is proposed. Experimental results indicate that the proposed grouped ciphertext feature recognition model achieves the highest accuracy in image classification and recognition, followed by normal text, with the lowest accuracy for random text. Nevertheless, the model’s classification accuracy for random text also exceeds 60%. Compared to the unimproved machine learning model, classification accuracy increased by 13% to 25%. Compared with other popular ciphertext feature recognition models, the proposed model demonstrates the best key sensitivity performance, with a maximum key sensitivity value of 1.6 and a minimum value of -0.9. Consequently, it is evident that the ciphertext feature recognition model developed in this research exhibits excellent key extraction and recognition performance, which can effectively enhance the information security of power network information systems.