The rise in cyber attacks has necessitated the creation of effective Intrusion Detection and Prevention Systems (IDPS). This work describes an improved approach to IDPS based on machine learning, using a Multi-Layer Perceptron (MLP) model for classification and anomaly detection. To ensure efficient model performance, the system uses a comprehensive dataset with one-hot encoded categorical characteristics and standardized numerical features. The model design includes four hidden layers with ReLU activations, followed by a softmax output layer that classifies network data as benign or malicious. The system is trained via backpropagation, with the Adam optimizer reducing cross-entropy loss. Extensive testing on a well-known intrusion detection dataset produces an excellent 97.3% accuracy, confirming the model’s capacity to accurately detect and prevent network attacks.