1 results listed
In this paper, a deep learning model has been developed to detect whether
malware/spyware leaks data to command and control servers and a new dataset has been
obtained from real-time environment for test of the model. In addition, effect of the size of the
data set and hyperparameters such as the number of layers of the deep neural network on the
success rate have been investigated. In this study, real-time data for harmful and normal İnternet
traffic have been obtained in the application layer and 100 features have been selected. The
developed deep learning model has been applied to 16,000 sample obtained from real-time
Internet traffic. From the experimental results, accuracy rates of 90% to 94% were obtained
with various number of samples and various number of layers in the deep learning model. It has
been seen from the experimental results that increase the number of samples increases the
accuracy rate. As well as, it has been seen that as increase the number of layers in the deep
neural network the accuracy rate increased first, further increase the hidden layers did not affect
the success rate. In this study, more distinctive and important features have been investigated
than others in the literature and the results have been tested.
International Data Science & Engineering Symposium
IDSES
Emrullah ERGİNAY
M. Ali AKÇAYOL