With the popularity of email as a significant communication tool in the digital environment, the emergence of cyber threats seeking to exploit email as an essential service comes through email scams commonly referred to as email spam that has proven to be a vector for more serious attacks such as phishing or malware distribution. This paper examines using several machine learning (ML) models in fighting against email-based threats and how they can bolster cybersecurity defenses. This study uses traditional models such as Logistic Regression and Decision Trees, as well as advanced algorithms, including Random Forests, Gradient Boosting, MLP, GRU, and LSTM, to generate a detailed analysis of each of these models concerning the accuracy, precision, recall, and F1 score. Furthermore, the performance of these models is optimized using Particle Swarm Optimization (PSO). Our results demonstrate the need to tailor cybersecurity frameworks to continuously adapt to the escalating cyber threat landscape by adopting advanced ML techniques.