Cybershield: Cyber Threat Detection System using Random Forest and MLP

Bhupathi Satpaksha

Computer Science and Engineering Hyderabad Institute of Technology

and Management Hyderabad, India bsatpaksha@gmail.com

K. Varunika

Computer Science and Engineering Hyderabad Institute of Technology and Management

Hyderabad, India varunika.koripoti@gmail.com

K.Ravi Kumar

Computer Science and Engineering Hyderabad Institute of Technology

and Management Hyderabad, India Kalligotla. ravikumar@gmail.com

K. Kranthi

Computer Science and Engineering Hyderabad Institute of Technology and Management

Hyderabad, India kranthipatel888@gmail.com

Darla Poorna Kala

Computer Science and Engineering Hyderabad Institute of Technology

and Management Hyderabad, India poornadarla04@gmail.com

K. Madhu Babu

Computer Science and Engineering Hyderabad Institute of Technology

and Management Hyderabad, India madhusurya3456@gmail.com

Abstract:The swift digitization of global infrastructure has resulted in a rise in challenges, as advanced cyberthreats increasingly target crucial and sensitive systems.To address this issue, CyberShield proposes a hybrid machine learning framework designed to predict and prevent cyberattacks before they occur. In order to detect complex network anomalies in real time, the framework combines Random Forest (RF) for ensemble-based feature evaluation with Multi-Layer Perceptron (MLP) for deep neural pattern recognition. The well-known benchmark datasets CICIDS2017 and UNSW-NB15, which provide realistic network traffic patterns and a variety of representations of modern attack behaviors, are used to train and validate the system. The hybrid architecture improves detection accuracy, lowers false positives, and exhibits strong adaptability to unseen attack patterns by combining these complementary techniques. According to experimental findings, CyberShield .The hybrid architecture improves detection accuracy, lowers false positives, and exhibits strong adaptability to hitherto unseen attack patterns by combining these complementary techniques, CyberShield outperforms individual models and traditional intrusion detection techniques, achieving detection accuracy of over 95%.This research highlights the importance of predictive learning methods in developing proactive cybersecurity measures and demonstrates their application in safeguarding digital infrastructures against evolving cyber threats.