An Optimal Modified Bidirectional Generative Adversarial Network for Security Authentication in Cloud Environment

Abstract

In the contemporary world, mobile appliances are connected to a cloud environment for balancing the wide range of data that is created. Because of the shortage of security and authentication, Cloud Computing (CC) is vulnerable to many attacks and security risks in the communication environment. Therefore, security is more significant for secure data transmission. In this paper, we proposed the modified Bidirectional Generative Adversarial Networks based on the Artificial Hummingbird (AH) algorithm (modified BiGAN-AH approach) within the cloud environment for security authentication. The ML-based BiGAN classifier is utilized to identify attacks during the session set-up stage. To reduce the loss, the cross-entropy function is computed in the training stage of BiGAN for estimating divergence. The discriminator’s parameters are predetermined while training both the encoder and generator. For enhancing encryption difficulty, the integration of the Substitution Permutation (SP) network and Feistel combination are used in the Symmetric Key Block Cipher (SKBC) algorithm. Data security is provided by the SKBC. To enhance the efficiency of GAN in detecting attacks, the Artificial Hummingbird (AH) algorithm is utilized for optimizing hyperparameters. The experimental outcomes reveal that the proposed approach reached 93.75% accuracy, 92.41% precision, 93.85% recall value, and a 93.52% F1 score.