An efficient detection of phishing sites in cloud computing using enhanced convolution neural network compared over linear regression with improved accuracy

The research aims to enhance the accuracy of phishing site detection in a cloud setting by employing innovative convolutional neural networks over traditional linear regression. The study involved comparing the performance of Novel Convolution Neural Network and Linear Regression algorithms using a sample size of 10, determined through a sample size calculation with a G-power of 0.8, alpha of 0.05, beta of 0.2, and a confidence interval of 95.52%. The Web page Phishing Detection Dataset, comprising 11,430 entries, was employed to identify phishing attacks, a prevalent method for acquiring confidential information from internet users. The findings revealed that the Novel Convolution Neural Network (95.52%) outperformed the Linear Regression (92.14%) algorithm significantly in detecting phishing sites. The Independent sample T-test exhibited a p-value of 0.003, below the significance level of 0.05, indicating a statistically significant difference between the study groups. In the context of cloud-based phishing site detection, Novel Convolution Neural Networks demonstrate superior accuracy compared to Linear Regression.