Article: Optimisation of generative AI models for cybersecurity threat detection through machine learning Journal: International Journal of Cloud Computing (IJCC) 2026 Vol.15 No.2 pp.257 - 279 Abstract: The optimising generative AI models, specifically variational autoencoders (VAE), generative adversarial networks (GANs), and tree-structured Parzen estimators (TPE), for enhancing cybersecurity threat detection. By integrating these models with traditional machine learning techniques, the framework generates realistic threat scenarios and simulates potential attack vectors, improving the detection of novel and unknown threats. The optimisation process fine-tunes generative models to create diverse datasets that reflect the full spectrum of cybersecurity threats, utilising ensemble methods and neural network architectures for better classification and identification. Evaluated using real-world datasets and simulation environments, the proposed approach demonstrated a remarkable 98.8% accuracy, with high precision at 98.2%, sensitivity at 98.5%, and specificity at 98.2%, alongside an impressive F1 score of 98.2%. The results show significant improvements over conventional methods, with reduced false positives. Future research should explore integrating these optimised models with real-time threat intelligence systems and broadening their application across various network environments. Inderscience Publishers - linking academia, business and industry through research

Title: Optimisation of generative AI models for cybersecurity threat detection through machine learning

Authors: Sarita V. Balshetwar; Girija Gireesh Chiddarwar

Addresses: Department of Computer Science Engineering, Faculty of Engineering, Yashoda Technical Campus, Satara, Maharashtra, 415011, India ' Department of Computer Engineering, Marathwada Mitra Mandal's College of Engineering, Pune, Maharashtra 411041, India

Abstract: The optimising generative AI models, specifically variational autoencoders (VAE), generative adversarial networks (GANs), and tree-structured Parzen estimators (TPE), for enhancing cybersecurity threat detection. By integrating these models with traditional machine learning techniques, the framework generates realistic threat scenarios and simulates potential attack vectors, improving the detection of novel and unknown threats. The optimisation process fine-tunes generative models to create diverse datasets that reflect the full spectrum of cybersecurity threats, utilising ensemble methods and neural network architectures for better classification and identification. Evaluated using real-world datasets and simulation environments, the proposed approach demonstrated a remarkable 98.8% accuracy, with high precision at 98.2%, sensitivity at 98.5%, and specificity at 98.2%, alongside an impressive F1 score of 98.2%. The results show significant improvements over conventional methods, with reduced false positives. Future research should explore integrating these optimised models with real-time threat intelligence systems and broadening their application across various network environments.

Keywords: cybersecurity; threat detection; variational autoencoder; VAE; generative adversarial networks; GANs; tree-structured parzen estimator; TPE; generative AI models.

DOI: 10.1504/IJCC.2026.154436

International Journal of Cloud Computing, 2026 Vol.15 No.2, pp.257 - 279

Received: 05 Nov 2024
Accepted: 05 Mar 2025
Published online: 29 Jun 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article

Title: Optimisation of generative AI models for cybersecurity threat detection through machine learning

Keep up-to-date