In cloud computing, there is an increasing threat from advanced attacks owing to increased network traffic and resource sharing. Conventional Intrusion Detection System (IDS) face challenges when dealing with datasets with high dimensionality and unbalanced classes. In order to solve these problems, propose a novel hybrid IDS that uses Deep Learning (DL), which combines TransSecureNet model with Dhole Optimization Algorithm (DOA). Data preprocessing consists of handling missing values and removing duplicates for providing quality and accurate input data. Synthetic Minority Over-Sampling Techniques (SMOTE) algorithm is utilized for creating synthetic instances and minority attacks classes to address imbalanced data issues. Feature engineering includes normalizing for making standardize data and enhancing training efficacy. TransSecureNet model combines a hybrid transformer encoder with Deep Neural Network (DNN) architecture for capturing temporal and non-temporal relations in the cloud network data. The DOA is used for tuning the hyperparameters, increasing the speed of convergence and improving the overall performance of the model. Experimental results carried out using the Python along with CIC-IDS 2018 datasets demonstrate the proposed model obtains an accuracy of 98.93%, a precision of 99.10%, a recall of 98.75%, and an F1 score of 98.92%, better than other existing approaches. The results indicate that the system of DOA-TransSecureNet delivers a stable and dependable solution for Intrusion Detection (ID).
Thankam et al. (Fri,) studied this question.