Email text analysis for fraud detection through machine learning techniques

Technology is improving and developing every day, new techniques and tools are releasing daily. Due to the development of technology, the number of new software and websites are growing every day. As the number of software is increasing the number of users using them is increasing as well. Hackers and malicious people will take this chance to do fraud, hack or trick especially the naïve users. Email has been the best way for communication for several fields such as education, business, and entertainment. Most companies rely on email for communication with customers or with other companies, due to this malicious people focus more on sending fraud emails or phishing emails. Even though people have become more aware of spam and fraud emails, still hackers are improving their email format and content to look like a legitimate email. In this research paper, four machine learning techniques are applied to detect illegitimate fraud email from legitimate email. Decision Tree, Random Forest, Naïve Bayes and Support Vector Machine classifiers are used for the experiments. These classification algorithms are applied on a new Fraud emails dataset consist of 11926 emails, where 5183 are Fraud (spam) emails and the rest are normal (ham) emails for their classification. The results show SVM has the best performance where the achieved accuracy is more than 98%.