Security is a main issue in Cyber-Physical Systems. Experts emphasize that Intrusion detection is a critical task for ensuring the security of computer networks. Detecting suspicious activities and generating alerts when an intrusion is detected is crucial for system monitoring. Various solutions are presented in the literature for intrusion detection but many anomalies and breaches are still a Problem for experts to eliminate the intrusion risk in Cyber-physical systems (CPS). The proposed method investigates the application of data normalization, principal component analysis (PCA), and neural networks for detecting attacks on the KDD99 dataset. It is clearly observed that dimensionality reduction using PCA is effective for this dataset. Further Neural Network model is used as an attack detection module on the transformed data. The proposed method achieves high accuracy in detecting both known and unknown attacks while reducing false positives, consistent with prior research. The proposed solution has practical implications for enhancing intrusion detection systems’ effectiveness and improving computer network security and increase the accuracy 97%.