Microsoft and Intel Transformed Malware Into Concrete Images

Microsoft and Intel Transformed Malware Into Concrete Images

Microsoft and Intel Transformed Malware Into Concrete Images

Microsoft and Intel’s research center Intel Lab is working on STAMINA, a new deep learning approach that will detect and classify malware. Tests for STAMINA show that the deep learning algorithm is highly successful.

Technology companies have long used machine learning to detect malware that targets users’ computers. Microsoft, one of these companies, is working with Intel on a new project to detect malware.

The deep learning project that Microsoft and Intel collaborated on was called STAMINA. STAMINA converts malware samples into grayscale images and compares the image it converts with textural and structural patterns of malware samples.

Working principle of STAMINA

Microsoft and Intel Transformed Malware Into Concrete Images

The team of Intel and Microsoft researchers explained how STAMINA works and stated that the whole process consists of a few simple steps. The first step is to take an input file and convert it into a one-dimensional pixel data stream.

The researchers then converted the generated one-dimensional (1D) pixel stream into 2D photography so that normal image analysis algorithms can be examined. The size of the 2D image to be created from the malware was also specifically determined. The width of the image was determined by the size of the input file. The image height is dynamically determined by dividing the pixel stream by the specified image width.

Intel and Microsoft’s researchers turned the resulting image into a smaller size image after converting the raw pixel stream into a normal 2D image. The researchers explained that resizing the raw image eliminates the need for computing resources to work with images consisting of billions of pixels, and the process gets even faster. The researchers also noted that the conversion of images into small-sized images does not adversely affect the outcome of the classification process.

Microsoft and Intel Transformed Malware Into Concrete Images

After creating the images, the deep neural network (DNN) was trained. To train the deep neural network, Microsoft offered to investigate 2.2 million infected file samples. 60 percent of infected file samples offered by Microsoft were used to train the DNN algorithm. While 20 percent of the files were used to verify the DNN, the remaining 20 percent were used for the actual testing of the project.

After the tests, STAMINA developers made a statement. The developers have announced that STAMINA has achieved a 99.07 percent success in identifying and classifying malware samples. Only 2.58 percent false positives were found in this detection and classification study of STAMINA.

Microsoft has long been investing in machine learning

Microsoft and Intel Transformed Malware Into Concrete Images

STAMINA research, developed in conjunction with Intel, can be seen as part of Microsoft’s efforts to improve malware detection processes using machine learning techniques. In the statement made by Microsoft about STAMINA, it was stated that STAMINA is not only correct and fast when working with small files, but also works with large files and can show its talents. However, it was underlined that STAMINA is less effective in small files than small ones.

Explaining Microsoft’s software security measures, Microsoft Threat Protection Security Research Director Tanmay Ganacharya explained that Microsoft now relies heavily on machine learning to detect emerging threats. Ganacharya also stated that Microsoft’s security system consists of different machine learning modules.

According to the announced results, STAMINA may be one of the machine learning modules that Microsoft uses to detect malware. Microsoft will make STAMINA work better with hundreds of millions of data it receives thanks to Windows Defender.

0 0 0 0 0 0
  • In-Site Comments

At least 10 characters are required

Next content:

Muscle production can be achieved without exercising with a new gene therapy

Article submission welcome to our system

Gallery Area

828 x 478