Volume 16, Number 1

Optimizing Hyperparameters for Enhanced Email Classification and Forensic Analysis with Stacked Autoencoders

  Authors

Merly Thomas¹ and B. B. Meshram², ¹Fr. Conceicao Rodrigues College of Engineering, India, ²Veermata Jijabai Technological Institute Mumbai, India

  Abstract

Electronic mail, commonly known as email, is a crucial technology that enables streamlined operations and communications in corporate environments. Empowering swift and dependable transactions, email is a driving force behind heightened productivity and organizational effectiveness. However, its versatility also renders it susceptible to misuse by cybercriminals engaging in activities such as hacking, spoofing, phishing, email bombing, whaling, and spamming. As a result, effective and efficient data analysis is important in avoiding and detecting cyber-attacks and crime on times. To overcome the above challenges, a novel approach named Aquila Optimization (AO) is used in this paper to find the best set of hyperparameters of the Stacked Auto Encoder (SAE) classifier. The purpose of increasing the hyperparameters of the SAE using the AO is to obtain a higher text classification accuracy. Then the optimized SAE classifies the selected features into different classes. The experimental results showed that the proposed AO-SAE model outperforms the existing models such as Logistic Regression (LR) and Long Short-Term Model based Gated Current Unit (LSTM based GRU) in terms of Accuracy.

  Keywords

Aquila Optimization, Cybercrimes, Email forensic dataset, ReliefF algorithm, Stacked Auto Encoder