This research investigates the application of Machine Learning (ML) techniques in two distinct security contexts: phishing detection and intrusion detection systems (IDS). The study demonstrates the effectiveness of ML models in classifying anomalies and patterns based on user-defined rules, significantly reducing the need for human intervention in data analysis. The methodology involved data collection, cleaning, feature selection, ML algorithm selection, model training, and fine-tuning. Results indicated a 90% ACC in phishing detection and 95% precision in IDS, with Decision Tree (DT) and Random Forest (RF) emerging as the best classifiers. However, challenges included data availability, cleaning, and variable selection requiring domain expertise. The evolving nature of security threats also affects the accuracy of supervised models. Future work involves deploying ML models in production for continuous learning and exploring advanced techniques like Adversarial Learning for improved classification. © 2025 Elsevier B.V., All rights reserved.