Skip to content

Better protection: AI in Cybersecurity

AI/ Machine Learning with Human Feedback for Cybersecurity

With the increase in Cybersecurity attacks such as ransomware, data breaches, business email scams, companies are spending more to protect their systems. AI used in addressing cybersecurity issues will allow increase in organizations’ ability to precent and respond to the ever-increasing cybersecurity attacks. Algorithms (i.e. modules) in AI systems can reduce time and effort in achieving cybersecurity tasks which are:

(1) Identifying and/ patching vulnerabilities (2) Detecting attacks, and defending against active attacks 1

AI/ Machine Learning with Human Feedback for Cybersecurity

An example of a cybersecurity system built on AI/Machine learning with human expertise approach (resulting in a predictive model of cyber-attacks) is depicted above (see figure above). This system starts up with the system (the unsupervised learning algorithm (i.e. module) within the system) monitors the network activity logs for indicators of malicious behavior (i.e. events); these indicators are ranked based on specific criteria and sent over to a human cybersecurity analyst. This analyst investigates and labels these events as either ‘normal’ or a type of specified attack. These labeled events are then sent back to the supervised learning module. This module them generates a model that is used to predict when an attack could occur in the future. In other words, there is recurring feedback between the human analyst and continually updated supervised learning module occurs. 2

This approach definitely improves performance in cybersecurity applications via detecting more events.

Reference

  1. US Government Accountability Office, “Artificial Intelligence Emerging Opportunities, Challenges, and Implications”, March 2018, GAO-18-142SP, https://www.gao.gov/products/gao-18-142sp ↩︎
  2. (1) ↩︎