An Explainable Artificial Intelligence Approach for a Trustworthy Spam Detection

For decades, spam emails have been one of the most serious and irritating cybersecurity threats. For detecting spam emails, a variety of machine learning (ML) and deep learning (DL) approaches are used. These approaches identify spam emails in the inbox and send them to a junk folder. However, these approaches have some limitations, such as their inability to explain why an email is considered spam. The current paper introduces the X_SPAM approach by combining the machine learning technique (Random Forest) and deep learning technique (LSTM) to detect spam and uses the Explainable Artificial Intelligence technique (LIME) to increase the trustworthiness of spam detection by explaining the reason for their classification. We evaluate the proposed approach using two different datasets (LING without metadata and Enron with metadata). We found that the proposed approach has achieved a high accuracy rate for RF and LSTM at 98.13% and 99.13% respectively. Moreover, the study exhibits a visualizing manner to eliminate the black box drawback for ML and DL classifiers to increase the approach's trustworthiness.