Optimized Deep Learning Model for Phishing Detection in Blockchain Transactions Using BERT and Teaching Learning-Based Algorithm

Phishing attacks in blockchain transactions present a serious threat to user security, exploiting fraudulent addresses and smart contracts. In this context, this work presents an improved deep learning model for phishing detection using Teaching Learning-Based Optimization (TLO) for hyperparameter tuning and BERT for feature extraction. The accuracy of our purposed model is 99.9% and it exceeded conventional models like RNN, LSTM, and GRU. The confusion matrix turned up just three false negatives and no false positives. Highly successful for phishing detection, the suggested approach offers a scalable and dependable way to improve blockchain transaction security against phishing assaults in practical uses.