Recognition of Abnormal Traffic Using Deep Neural Networks and Fuzzy Logic

Recognition of the abnormal traffic using the deep learning is proposed. The dual deep neural network architecture for detecting the DOS attacks based on a combination of convolutional layers is proposed. The new thing is the classification strengthening realized by the reinforcement of the input vector with its cluster evaluation. The software was implemented, the testing bench was assembled and a semi-natural experiment was conducted on the basis of the local computer network of the University. As a result of the tests, 87% accuracy of attack recognition was achieved.