Email Filtering based on Artificial Neural Network: Activation functions and Learning rate analysis

The information and communication technologies evolution, along with the development of the Internet, has meant that networks and information systems now play a crucial role in the society, whose citizens security is not marginalized. Most successful cyberattacks start with a well-known vector, which is email using social engineering techniques as spam and phishing. Email attacks are more and more frequent, therefore it is necessary to know how we can protect the confidential information. This paper aims to detect phishing and spam emails using high-accuracy machine learning techniques using the ANN algorithm with data preprocessing, while changing the learning rate for each activation function to find the one that gives the minimum error in terms of RMSE for each function.