An Efficient Feature Selection-based DDoS Attack Detection in Cloud Computing using Optimized LSTM

M.C Malini, N. Chandrakala

Abstract: DDoS attacks are considered one of the most severe security risks to the cloud computing environment by the fact that it is capable of overloading resources, affecting the service availability. The dimensionality, redundancy, and time constraints of cloud network traffic are quite high, which complicates the use of traditional intrusion detection systems. To resolve these problems, the proposed paper will propose an effective DDoS attack detection model, which combines autoencoder-based feature selection and an optimized Long Short-Term Memory (LSTM) model. The autoencoder is used to learn in an automatic manner compact and discriminative feature representations of high-dimensional traffic data, and thus eliminate redundancy and enhance learning efficiency. In order to improve the performance of detection, the LSTM network is optimized by an Improved Firefly Algorithm (IFA), which is augmented by Partial Opposition-Based Learning (POBL). Diversification of the population is enhanced by the use of POBL, which also speeds up convergence, allowing a good tuning of hyperparameters without premature convergence. The optimized LSTM is very effective in capturing long-term temporal dependencies in the network traffic, which are necessary in the correct differentiation of DDoS attacks and normal cloud traffic. The proposed framework is tested on benchmark DDoS datasets frequently utilised in cloud security studies, and the performance of the framework is compared with traditional LSTM and alternative metaheuristic-optimised LSTM frameworks. The experimental findings indicate that the suggested method has high accuracy, precision, recall, and F-score, as well as a faster and more stable convergence rate. The results substantiate that the autoencoder-selected features, combined with IFA, POBL-optimized LSTM, give a solid, efficient, and scalable algorithm to detect DDoS attacks in real time in the context of cloud computing.

Keywords: DDoS attack detection, hyperparameters optimization, firefly algorithm, long short-term memory, partial opposition-based learning, autoencoder

How to Cite?: M.C Malini, N. Chandrakala, "An Efficient Feature Selection-based DDoS Attack Detection in Cloud Computing using Optimized LSTM", Volume 14 Issue 12, December 2025, International Journal of Science and Research (IJSR), Pages: 2288-2300, https://www.ijsr.net/getabstract.php?paperid=SR251226161456, DOI: https://dx.doi.org/10.21275/SR251226161456