Quantile Regression Neural Network with Rhinopithecus Swarm Optimization Algorithm and Paillier Encryption for cyber security enhancement
S. Sathyakala, E. Anbalagan · 2025
Growing dependence on digital technology has made cybersecurity a more significant area of study and application. Its primary focus on safeguarding equipment, networks, information and other resources against various cyber-attacks, threats, risks, damages, or unauthorised access. However, creating a successful AI-based security model these days is difficult due to the dynamic and complicated nature of real-world scenarios. Furthermore, there is still an issue in effectively defending against various attacks. To overcome these issues, Quantile Regression Neural Network (QRNN) with Paillier encryption is developed to enhance the network security. Initially, cyber threat data is collected and pre-processed using Robust Scalar Normalization (RSN), and missing values are imputed through a Regression based Missing Value Imputation (RMVI). Then Barnes-Hut t-Distributed Stochastic Neighbour Embedding (BH-t-SNE) is employed which reduces the data's dimensionality. After that, QRNN is utilized to detect cyber-attacks, and in order to enhance the performance of the QRNN classifier, a number of quantiles and decay factor are optimally selected using the Rhinopithecus Swarm Optimization (RSO) algorithm. Finally, a Paillier Encryption algorithm (PEA) is used to securely store data in the cloud server. The proposed approach has an accuracy of 98.1%, and PPV of 92.40%. By integrating deep learning with cryptography creates a smarter, more secure, and privacy-preserving defence system for detecting and mitigating cyber threats.