Robust Phishing Detection in Consumer IoT Devices with ANOVA F-Test and Satin Bowerbird Optimization of Deep Learning Model

Combining ANOVA F-Test for feature selection with Satin Bowerbird Optimization (SBO) for hyperparameter tuning of a deep learning model, our proposed model delivers a strong phishing detection model for consumer IoT devices. In terms of accuracy and loss, the suggested CNN model, optimized using SBO, outfits GRU, LSTM, and RNN models. With a significant loss decrease, the model had a high accuracy of 92% and proved effective in spotting phishing attempts. Comprehensive assessments including feature selection, correlation analysis, and performance comparisons indicate the model's excellence in both training and testing stages, therefore providing an efficient means of improving security in IoT systems. This solution offers a scalable and effective means of phishing detection for smart home appliances.