Many companies are puzzled by the development of tools to detect malicious software. This task is a priority problem of cybersecurity for the whole society. Some types of malware can inflict millions of dollars of damage.
The article presents an artificial neural network, trained to distinguish between benign and malicious executable Windows files with only the original byte sequence of the executable file as input data. This approach has a number of practical advantages.
In conclusion, this document shows that neural networks are able to learn to recognize benign and malicious executable files of Windows without costly and unreliable development of functions.