Spam emails represent a formidable cybersecurity threat, necessitating precise classification methods to mitigate associated risks and reduce the influx of unwanted messages. This research delves into the quest for improved spam email classification accuracy by leveraging ensemble machine learning techniques, specifically focusing on the utilization of origin, content, and image features within emails. Our study involved the training and testing of a random forest classifier, assessing individual features and integrated features using metrics such as accuracy, recall, precision, and F1 score. The results revealed that the hybrid framework, which combines these features, outperforms individual feature-based approaches. In the context of the ever-evolving landscape of spam technology and the emergence of novel message types challenging traditional methods, we introduce an integrated approach. This approach integrates feature results from various sub models to achieve superior classification accuracy. Our findings demonstrate the outstanding performance of the hybrid approach, achieving the highest accuracy rate (97.6%), recall rate (95.9%), precision rate (98.9%), and F1 score (97.4%) among the tested techniques. The research presents an innovative amalgamation of features that significantly enhances classification accuracy, making a notable contribution to the existing body of knowledge. It underscores the importance of feature integration technique in the field of spam email classification.