MUMBAI, India, Nov. 2 -- Intellectual Property India has published a patent application (202641127109 A) filed by Karpagam Academy Of Higher Education; Karpagam Institute Of Technology; Dr. S. Mythili; and Rahul R on October 26, 2026, for Cyber Threat Attack Level Detection Using Machine Learning.
Inventors include Dr. S. Mythili; and Rahul R.
The application for the patent was published on October 30, 2026, under issue no. 44/2026.
Abstract: The rapid growth of digital services and interconnected systems has intensified the frequency and sophistication of cyberattacks, making real-time threat detection a critical requirement for modern organizations. Traditional rule-based and signature-driven security mechanisms are increasingly inadequate against dynamic and multi- stage attack patterns, leading to delayed responses and high false-positive rates. To address these challenges, this study presents a machine learning–based cyber threat attack level detection system designed to enhance accuracy, adaptability, and proactive threat management in contemporary cybersecurity environments. The proposed system continuously monitors heterogeneous security data sources, including user behavior logs, network traffic flows, and system-generated event records. By integrating these multi-source data streams, the system achieves comprehensive situational awareness and enables effective detection of anomalous activities that may indicate potential cyber threats. Supervised machine learning classifiers are employed to learn complex behavioral and traffic patterns, allowing the system to identify both known and previously unseen attacks with high precision. A key contribution of this work is the introduction of a learning automata–based malicious social bot detection algorithm, referred to as LA-MSBD. The algorithm dynamically adapts its decision-making strategy based on feedback from the operating environment, ensuring continuous learning and improved robustness against evolving attack behaviors. To enhance detection reliability, the system incorporates a hybrid trust evaluation framework that combines direct and indirect trust assessments. Direct trust is computed using Bayesian probabilistic inference derived from real-time observations, while indirect trust is evaluated using Dempster–Shafer evidence theory to manage uncertainty and conflicting evidence from distributed data sources. The fusion of these trust metrics significantly improves the identification of malicious entities that mimic legitimate user behavior. The system further classifies detected threats according to attack severity levels, enabling prioritized alert generation and efficient incident response. This attack level classification approach allows security administrators to focus on high-risk events while minimizing alert fatigue. The proposed framework effectively detects common web-based and network-level attacks, including SQL injection, distributed denial-of-service (DDoS) attacks, and cross-site scripting (XSS) vulnerabilities. Implementation of the system within a scalable Django-based web framework facilitates real-time deployment, modular integration, and adaptability to enterprise and cloud computing infrastructures. Experimental evaluations conducted on multiple benchmark network datasets demonstrate that the proposed approach outperforms existing methods in terms of precision, recall, F-measure, and overall accuracy. The system achieves a detection accuracy of 95.3%, highlighting its effectiveness in real-world cybersecurity scenarios. Overall, the proposed solution provides a scalable, adaptive, and intelligent mechanism for proactive cyber threat detection and attack level assessment in modern digital environments.
Disclaimer: Curated by HT Syndication.
