There are ways and ways of detecting the malware in the code of an application or file, but none of them as original as Microsoft and Intel. In an attempt to better automate the search for malicious code, the two companies have developed an artificial intelligence system that transforms the code into an image and searches it for the possible hidden virus.
Under the name of STAMINA (STAtic Malware-as-Imago Network Analysis), this technique seeks to facilitate the detection of malware by transforming all the code into grayscale photographs. With this, the AI has less “work” to do in the search as it is easier to find similar patterns in images than in code as such.
Microsoft and Intel take it From code to pixels
According to Intel, the process follows relatively simple steps. The binary code of the file to analyze is transformed into a sequence of pixel data and later into a two-dimensional image. To further reduce AI computations, this image is reduced in size so it doesn’t have to process billions of pixels. They explain that this reduction in the tests carried out has not affected the final detection result.
Once the image is ready, it happens to be analyzed by the AI. The AI has previously been fed malware samples from 2.2 million hashes of infected executables. From there the AI only has to find similar patterns and textures between the image (file) to analyze and the malware samples that it has stored.
Researchers at Intel and Microsoft say AI was able to identify and classify malware with 99.07% accuracy using this technique. Of course, with a false positive rate of 2.58%. According to Microsoft, the technique is accurate and fast for small files, although it can become less effective if you have to work with large files due to what it means to transform this into images.
👇 More in NUpgrade