MUMBAI, India, May 1 -- Intellectual Property India has published a patent application (202641050569 A) filed by Srinivasa Ramanujan Institute Of Technology; C. Ravi Teja; P. Sufiya; M. Muskan; and K. Supriya, Ananthapuramu, Andhra Pradesh, on April 21, for 'ai driven cyber threat detection and attack attribution using deep learning.'
Inventor(s) include Srinivasa Ramanujan Institute Technology; C. Ravi Teja; P. Sufiya; M. Muskan; and K. Supriya.
The application for the patent was published on May 1, under issue no. 18/2026.
According to the abstract released by the Intellectual Property India: "Cyber-physical systems are important to the modern industrial infrastructure of today, but their expanded connectivity opens them to more advanced types of cyberattacks. A serious obstacle in the detection of these attacks is that there is a large amount of data that is built on normal operational records and very few samples of malicious operations, very little amount of malicious records will create a model that will make it very difficult to detect a cyberattack with reliability and will cause a bias in the detection model. In this project, we will present a hybrid model for the detection and attribution of cyber-physical systems using deep learning and that does not rely on traditional over-sampling approaches or under-sampling approaches.The approach presented in this study involves using an AutoEncoder to extract features from imbalanced datasets. After extracting features from the data with Autoencoder, Principal Component Analysis (PCA) is applied to reduce the number of dimensions of the feature sets. Next, a Decision Tree classifier is applied to classify types of attacks based on the reduced feature sets.Finally, a Deep Neural Network (DNN) is trained to correctly identify both known and unknown types of attacks based on the previously supplied outputs of the Decision Tree. The proposed framework is tested using the Secure Water Treatment (SWAT) dataset, which consists of a variety of types of real-world cyber-attacks, including response injection, command injection and denial-of-service (DoS) attacks. Experimental results demonstrate that the proposed approach effectively addresses class imbalance while providing a scalable and generalizable solution for securing cyber-physical infrastructures through the integration of deep feature learning and hierarchical classification techniques."
Disclaimer: Curated by HT Syndication.
