The proliferation of the Internet and the fast development of technology, there naturally arises the abomination of cyberattacks that urge the network to become more secure. Typical intrusion detection systems (IDSs) are incapable of precluding very recent threats such as zero-day vulnerabilities and polymorphic malware. This paper delves into the adoption of machine learning (ML) as a means of improving IDS capabilities through the use of algorithms like Random Forest, Support Vector Machines (SVM), K-means clustering, and Deep Autoencoders. Using real-world datasets like NSL-KDD and CICIDS2017, and techniques like dimensionality reduction shows enhanced detection accuracy, scalability and flexibility. To protect the system from previously known epidemics, we used Random Forest. Autoencoders have shown the key technology in anomaly detection. By creating ML-based IDS System that secures network from any cyber attack and threats can be used as one of the realistic way in this research