Next-Gen Phishing Detection System Based on Federated Learning Integrated CNN-LSTM for SMS Communication

Phishing is one of the most common types of cybercrime. It is a method used by attackers to reveal personal information. The possibility of malicious activity is increasing in the rapidly changing the internet, where technological improvements enable seamless connectivity and information exchange. Conventional phishing detection techniques find it difficult to detect the dynamic strategies used by the malicious actors. This research study makes use of a hybrid Convolutional Long S hort-Term Memory (ConvLS TM) neural network to present a novel method for improving the detection rate of phishing in text-based communications. The core of the proposed system lies in combining the ConvLS TM architecture with a Federated Learning (FL) framework. The FL tackles the privacy and data decentralization challenges. This makes it possible to train the model on a variety of entities or organizations. The federated hybrid ConvLS TM model achieves higher accuracy, precision, and recall rates than conventional phishing detection techniques, as shown by the testing results. The proposed model can generalize well over a wide range of datasets from various organizations by utilizing the federated method, which further makes it a reliable and useful tool for detecting phishing attacks in an era where data safety and accuracy plays a major role.