VC DeepFL: Vector Convolutional Deep Feature Learning Approach for Identification of Known and Unknown Denial of Service Attacks

Nowadays Denial of Service (DoS) attack is one of the threatening cyber attacks for the users of Internet which denies online services for legitimate users. Therefore, DoS attack detection mechanism is needed to protect the online services from these attacks. A number of machine learning based attack detection mechanisms exist and the existing detection mechanisms face the lacuna of identifying known and unknown DoS attacks and suffer from low accuracy and high false alarm. In this paper, these issues are addressed by proposing a Vector Convolutional Deep Feature Learning (VCDeepFL) approach for identification of DoS attacks. The VCDeepFL approach is a combination of Vector Convolutional Neural Network (VCNN) and Fully Connected Neural Network (FCNN). VCNN extracts the feature by down sampling the input vector and provides a better representation of input vector. FCNN is a multiclass classifier that boosts the performance of the attack detection system by automatically computing the best set of weights from training. The proposed approach is tested with NSL KDD dataset and compared with state of the art attack detection system and base classifiers. It is evident that the proposed approach yields prominent results for most of the classes. Further, Receiver Operating Characteristics (ROC) analysis is performed and it is seen from the ROC curve that the Area Under the Curve (AUC) is high for the proposed approach.