Machine Learning for malware detection

Module 10 of C1b3rWall Academy 2021, “Forensic analysis. Malware and stealth techniques”, moves to the second block of content. It begins with a keynote lecture by Mario Guerra from Disruptive Consulting, which focuses on reducing, as much as possible, the advantage cybercriminals have in terms of their malware analysis capabilities. Mario explained that, in threat detection, there are several categories such as configuration analysis, threat modeling, indicators of compromise and the behavior of the threat itself. These categories relate to the following variables:

  • Transparency. Visibility of an alert that analytics brings to the analyst (what happened and what it means).
  • Durability. Readiness with which the analytic approach withstands adversary change.
  • Ease of implementation. Amount of time, effort, and data required for implementation.
  • Coverage. Number of analytics needed to provide maximum research coverage.
  • Adaptability. How well an organization tunes the approach to its specific environment to improve the false positive/negative rate.

Machine learning-based detection

He also explains that everything to do with machine learning has its origin in big data. The goal is for the system to be able to detect harmful agents on the basis of the criteria established in the algorithm. The creation of personal algorithms makes it possible to adjust them to special needs, achieving a better or more specific product than a commercial product, making it possible to reduce false positives/negatives.

When samples of malware are available from a threat group that is relevant to the organization, the development of detection technologies based on machine learning can aid in the detection of new samples of such threats.

The difference between a machine learning algorithm and a traditional one is that the latter provides the system with sequential step-by-step instructions on how to operate, whereas machine learning systems learn from the samples to differentiate between good and bad according to the defined threshold. This will make it possible to automate the work of signature creation.

We still have a lot to learn about cybersecurity, but every subject that our speakers touch on is very interesting. I invite you to follow the complete program of the C1b3rWall Academy 2021, which I assure you will not be in vain.

At the following link is the complete News-365 article and you can register here to access all the presentations belonging to this module.

Juan Manuel Corchado

Full Professor in Computer Science and Artificial Intelligence, Department of Computer Science and Automation, University of Salamanca, Spain.


Contents: – e4YOU – Másteres – CyberCamp

Scientific: ORCIDScholarResearch GateWeb of Science

Join 8,703 other subscribers