| Abstrakt: |
In cloud environment, Denial-of-Service (DoS) and Distributed Denial-of-Service (DDoS) attacks poses a major challenge to the accessibility. To resolve this issue, an intrusion detection system (IDS) can be employed as a security procedure which operates on the network layer. The conventional IDS in cloud platform result to poor detection accuracy with high computational complexity. Keeping these issues in mind, in this paper, we introduce an efficient feature subset selection based classification model for the identification of DDoS attack. To detect the DDoS attack in IDS, best feature sets are selected with maximum detection by the use of Random Harmony Search (RHS) optimization model. Once the features are selected, a Deep learning based classifier model using of Restricted Boltzmann Machines (RBM) is applied to detect the DDoS. For improving the detection rate of DDoS attack, a set of seven extra layers are included among the visible and the hidden layers of the RBM. Here, precise results are attained by optimizing the hyper parameters of the presented deep RBM model. The probability distribution of the visible layer in RBM model undergoes replacement with a Gaussian distribution. For experimentation, RHS-RBM model is tested against KDD′99 dataset. The experimental results showed that the RHS-RBM model achieves maximum sensitivity of 99.88, specificity of 99.96, accuracy of 99.92, F-score of 99.93 and kappa value of 99.84. These obtained values of RHS-RBM model are found to be better when compared to the RBM model without the use of RHS algorithm. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink