User Identification based on Behavior Feature Analysis

The issue of identity management and information security in cyberspace is becoming increasingly important, and the issue of how to effectively identify through user behavior urgently needs to be addressed. At present, biometric recognition technologies have certain limitations. This article combines deep learning methods and uses the BERT model to extract input text information features to generate word vectors. It uses a bidirectional GRU network with local and global attention mechanisms for training, and finally uses a softmax classifier to classify the user identity recognition model. The model in this article has significantly improved its performance in user identity recognition tasks based on behavioral features.