International Journal of Science, Engineering and TechnologyAuthors: Ms.Monika Saini, Dr. Gaurav Aggarwal
Abstract: This research presents a novel hybrid machine learning (ML) framework, SART (Supervised-unsupervised Anomaly Recognition Threat detection), for enhancing cyber security by integrating supervised, unsupervised, and deep learning techniques. The study addresses the limitations of traditional cyber security methods by leveraging advanced ML algorithms to improve threat detection, anomaly identification, and predictive defense mechanisms. Using benchmark datasets such as CICIDS2017, UNSW-NB15, KDDCup99, CSE-CIC-IDS2018, and NSL-KDD, the proposed framework combines Random Forest (RF), Support Vector Machine (SVM), Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) networks, and Auto encoders to achieve high detection accuracy and robustness against evolving cyber threats. The framework integrates security mechanisms including AES encryption, RSA asymmetric encryption, and SHA hashing for confidentiality, authentication, and integrity. The experimental results demonstrate significant improvements in security, accuracy, and performance compared to conventional approaches, highlighting the effectiveness of the hybrid ML-based model in real-world cyber security applications. The proposed model achieved detection accuracy exceeding 97% on benchmark datasets while maintaining computational efficiency suitable for real-time deployment.
DOI: http://doi.org/10.5281/zenodo.18093494
