Email Security Enhancement via ML-based Spam Filter Techniques

Because email correspondence is significantly increasing these days, there arises a high percentage of spam email, resulting in security risks from phishing to the distribution of malware and thefts of identities. The focal intention of this research is the formulation and comparison of machine learning based spam-detecting methods to provide accurate email classifications such as either spam or not spam. This paper experimented on three classifiers namely Support Vector Machine, Random Forest and Naive Bayes. Support Vector Machine is a strong supervised algorithm that finds out the best decision boundary for the classification process. Naive Bayes is one of the probabilistic models derived from Bayes theorem that is efficient for text classification. Random Forest is a method of ensemble learning, attempts to enhance the classification accuracy by averaging ensembles of decision trees. The precision, recall, accuracy, and F1-score of such models are compared using a public dataset of spam emails. The trades and constraints of each algorithm will be indicated, providing some insights into whether these algorithms may be useful for real-world applications of email filtering. The work here compares the efficiency of Random Forest, Naive Bayes and SVM machine learning models by the use of metrics precision, recall, F1-score and accuracy and also identifies the effective one in the detection of spam and eventually to better email safekeeping and user satisfaction.