Improving Phishing Detection Accuracy Through Feature Selection and Deep Learning Model Integration

MEKALA MANOJ KUMAR
Department of computer science and engineering,
Vemu Institute of technology
kothakota, chittor district,
Andhra Pradesh 517112,
Mailid: mekalamanoj7359@gmail.com

Mr. K. Niranjan
Assistant Professor Vemu Institute of Technology,Dept of CSE

Mail id- kalikiriniranjan@gmail.com

Abstraction: Phishing attacks are a persistent cybersecurity threat, and thieves are able to steal important information by exploiting people's trust. In this paper, we present an enhanced phishing detection framework, which is based on the combination of advanced feature selection methods and Machine Learning and Deep Learning algorithm. The paper leverages the use of labeled dataset with each instance being either legitimate or phishing and introduces the performance of using different models Graph Convolutional Networks (GCN), TabTransformer, Autoencoders, Feedforward Neural Networks (FNN) and Deep Neural Networks (DNN). Finally, the feature selection process should be optimized to help improve the model accuracy, decrease the computational overhead, and improve the generalization. The solution has been coded in python and released in the form of a flask web application with a simple HTML and CSS based user interface. Our experiment results support the conclusion that a deep learning architecture in combination with a proper feature selection can facilitate a better performance and robust phishing detection mechanism. This is a real-life implementation model which can be used to achieve a scalable and effective phishing control.
Keywords: DNN, FNN, Autoencoder, GCN, TabTransformer,Featuregeneration,Detection, Flask, Cybersecurity.