With the progressive proliferation of cloud computing applications in recent decades, the emergence of novel challenges within this domain is inevitable. Among these challenges, security constitutes a fundamental prerequisite for the effective utilization of such environments. A pivotal concern in the domain of cloud computing security is intrusion detection, which, if inadequately addressed and lacking sufficient accuracy, can precipitate risks such as data breaches and potential misuse. Consequently, precise intrusion detection and the enhancement of protective system efficiency assume particular significance. Over the years, scholars have diligently endeavored to optimize this challenge. Nonetheless, numerous extant methodologies have encountered the issue of relatively low detection accuracy. A contributing factor to this dilemma is the heterogeneity among irrelevant and redundant informational features present within the datasets. This article introduces an innovative solution for intrusion detection within cloud infrastructure, leveraging an enhanced neural network classification methodology grounded in Grey Wolf Optimization (GWO). The objective of this approach is to augment detection accuracy. In this context, the principal component analysis (PCA) algorithm has been initially employed to mitigate the dimensionality of the dataset. Subsequently, by employing artificial neural network classification, normal traffic is distinguished from malicious traffic. Following this, the parameters of the artificial neural network undergo optimization via a Grey Wolf Optimization. Ultimately, the performance of the proposed methodology is assessed and evaluated using the standard and comprehensive NSL-KDD dataset, serving as a benchmark for evaluating intrusion detection systems, within the MATLAB environment and compared against the outcomes derived from three other prominent methodologies. The results of this comparative analysis indicate that the proposed methodology exhibits superior performance, achieving a detection accuracy of 99.41% relative to alternative methods, and demonstrates an enhanced capability for identifying attacks with greater precision. Clinical Trial Number: Not applicable.
Farahi et al. (Fri,) studied this question.