This study proposes a hybrid cyber security approach, combining signature-based and anomaly based detection systems to identify packet-based attacks effectively. By utilizing machine learning algorithms such as Decision Trees, Random Forest, and Gaussian NB, the system aims to improve detection accuracy and efficiency. Signature-based detection relies on known patterns, while anomaly based detection identifies deviations from established norms. The research evaluates the performance of these algorithms in detecting network packet attacks through comprehensive experimentation and analysis. Results highlight the strengths and limitations of each model in accurately pinpointing various types of attacks. The findings emphasize the significance of a hybrid approach, utilizing both signature and anomaly detection, and showcase the potential of machine learning algorithms in bolstering network security against evolving cyber threats.