Detection of Malware with Deep Learning Method

Abstract

Nowadays, many studies are done on the detection of malicious software. Static, Dynamic and Hybrid analysis methods are used to collect data for malware detection. With these methods, data is created by reading the information in the file without running the malicious software, or by examining the places it affects such as changes on the network at runtime, api calls. With the advancement of today’s technology, these data are combined with Machine learning algorithms or architectures of Deep Learning to detect malware. Detection of malicious software On the data set containing malicious software, it was detected by using CNN and ANN neural networks. While close to 10,000 datasets showed a success rate of close to 99%, datasets close to 50,000 achieved close to 97% success. In our study, a success rate of 98.1% was achieved for nearly 50,000 data sets. Among the studies researched, malware detection was made with higher accuracy than the studies using data sets closest to 50,000.