MUMBAI, India, July 24 -- Intellectual Property India has published a patent application (202621063821 A) filed by Dr. Vikas Sharma; Rukmani Sharma; Gopal Laharpure; Shashank Charpe; Payal Suryavanshi; Manisha Kushwaha; Pinky Kushwaha; Nishant Gurjar; Hekrishna Bhargava; and Hema Bhavane on May 20, 2026, for An Ai-Based Real-Time Cyber Attack Prediction And Prevention System Using Deep Neural Networks.
Inventors include Dr. Vikas Sharma; Rukmani Sharma; Gopal Laharpure; Shashank Charpe; Payal Suryavanshi; Manisha Kushwaha; Pinky Kushwaha; Nishant Gurjar; Hekrishna Bhargava; and Hema Bhavane.
The application for the patent was published on July 17, 2026, under issue no. 29/2026.
Abstract: ABSTRACT An AI-Based Real-Time Cyber Attack Prediction and Prevention System Using Deep Neural Networks The present disclosure relates to an intelligent and adaptive cybersecurity system that employs Artificial Intelligence (AI) and Deep Neural Networks (DNN) for real-time prediction and prevention of cyber-attacks in modern digital environments. With the rapid expansion of interconnected systems such as enterprise networks, cloud computing platforms, and Internet of Things (IoT) devices, the attack surface for cyber threats has significantly increased. Traditional security solutions, which rely on predefined rules and signature-based detection, are often incapable of identifying emerging and sophisticated attacks. The proposed invention addresses these limitations by introducing a data-driven approach that continuously monitors network traffic, system activities, and user behavior to detect anomalies and predict potential threats before they occur. The system operates through a multi-layered architecture that includes modules for data acquisition, preprocessing, feature extraction, deep neural network-based prediction, and threat classification. The data acquisition module captures high-volume, real-time data streams from various sources, which are then cleaned and transformed into structured formats through preprocessing techniques such as normalization and encoding. The feature extraction module identifies critical parameters, including traffic patterns, protocol behavior, and connection statistics, which are fed into the deep neural network. The DNN model, comprising multiple hidden layers and advanced activation functions, learns complex and non-linear relationships within the data, enabling it to accurately classify network activities into normal, suspicious, and malicious categories. This predictive capability allows the system to identify both known attack signatures and previously unseen threats, including zero-day vulnerabilities. In addition to detection and prediction, the invention incorporates an automated prevention and response mechanism that enhances system resilience by taking immediate corrective actions upon identifying a threat. These actions include blocking malicious IP addresses, terminating unauthorized sessions, isolating compromised nodes, and dynamically updating firewall and security policies. Furthermore, the system integrates a feedback-based learning framework that continuously retrains the deep learning model using newly generated data, thereby improving its accuracy and reducing false positive and false negative rates over time. The proposed system offers a scalable, efficient, and proactive cybersecurity solution that can be deployed across diverse domains such as financial institutions, healthcare systems, government infrastructures, and smart city environments, ensuring robust protection against evolving and sophisticated cyber threats.
Disclaimer: Curated by HT Syndication.
