Email spam is a significant problem in information technology because it can disrupt productivity, steal sensitive information, and damage the reputation of email systems. This study implemented the Gaussian Naive Bayes (GNB) algorithm to classify spam and non-spam (ham) emails. The GNB algorithm was chosen because it is capable of handling continuous data and has a simple training and prediction process. The dataset used consists of 5,136 emails divided into two categories: 4,496 ham emails and 640 spam emails. To address data imbalance, the Random Oversampling (ROS) method was applied by duplicating data in the minority class so that the number of datasets increased to 8,992. The results showed that the model achieved an accuracy rate of 94.30%. In the ham class, the model produced a precision of 100%, a recall of 87.97%, and an F1-score of 93.60%. Meanwhile, in the spam class, the precision was 90.24%, a recall of 100%, and an F1-score of 94.87%. These results demonstrate that the Gaussian Naive Bayes algorithm is effective in detecting spam emails with a high degree of accuracy. However, there are still some misclassifications of ham emails, so further development is needed to improve model performance, particularly in reducing misclassifications of non-spam emails