Deep Learning Model with Optimization Strategies for DDoS Attack Detection in Cloud Computing

Cloud Computing (CC) remains as a chief research area for analysts because of its numerous applications and advantages. The dispersed nature of CC and its dependence on internet service poses several attacks and challenges. Mainly, the attacks like DDoS attempts to disrupt the online operations. Insider attacks cannot be detected using conventional detection techniques like firewalls. This study suggests a DDoS detection method for reducing the DDoS activity. The proposed work contributes under big data perspective by handling certain procedure for detection process. Data Generation phase is the initial step, and according to the work, DDoS dataset is considered under big data perspective. As the work is considered under the big data perspective, Map reduce (MR) framework is used, Mapper handles the data and process the feature extraction, which includes the extraction of raw features, Packet feature extractor, Improved Correlation and statistical Feature. Reducer provides the combined feature set. From the extracted feature set, appropriate features are selected via Weightage based Improved hybrid model combining the models like Long short term memory(LSTM) and Deep Maxout(DMO) networks is used. The weights of LSTM and DMO are optimally chosen using White Shark-Remora Optimization (WSROA) algorithm. Recursive Feature Elimination (RFE) process is used to precisely select the features for DDoS attack detection. The findings shows that the proposed hybrid bio-inspired algorithm outperforms with an accuracy of 94 % compared to LSTM, Convolutional Neural Networks(CNN) and Deep Belief Networks(DBN)