Designing An Intelligent Architecture to Detect Malevolent E-Mail Using Machine Learning Approach

The most common method of remote communication in business, corporate, and personal life is email. It contains a vast amount of private data that is sent across the network. So, these mails have to be managed in an efficient manner for maximal and uninterrupted communication. Unfortunately, spam-a term for malevolent emails-many a time enters our e-mail accounts due to an individual's dishonest intentions. Spam puts more load on ISP servers and bandwidth, and consumers are responsible for bearing the additional costs incurred in handling this load. Spam emails are designed to send unsolicited commercials and useless proposals to specific people; therefore, it has been subject to laws in several jurisdictions. Consequently, the categorization of authentic emails from a large email dataset is the main intent of this study. Both synthetic data created from many email accounts and regular datasets gathered from the UCI Machine Learning Repository were used in the experiments. Subsequently, data was imported from multiple sources, and data transformation technique was applied to the dataset. "Principal Component Analysis" is used as an attribute selection strategy where significant attributes are identified. This paper describes classification of E-mails by Random Forest (RF algorithm). The improved dataset after PCA has been classified by "Random Forest classification algorithm" which outperforms large datasets. Plot Layout, Receiver Operating Characteristics (ROC) graph, and bar chart visualization techniques are used to report the results.