Protecting against APTs with Machine learning

Machine learning is a science that uses existing data on a subject to train a computer how to identify related data.  Just like with humans, the more training a machine learning algorithm gets, the more likely it is to succeed at its task.  We have an extensive amount of information on attacks that can be used to train machines.

After all, new attacks come out every day and over a hundred million malware samples have been collected each year since 2014.  This information, as well as the historical information, can be fed into machine learning algorithms to better understand the attacks that haven’t happened yet.  Machine learning systems are comprised of algorithms that determine how the program will interpret, understand, and correlate information to make decisions.  As new data is added to a machine learning system, it can produce results which are tested and then refinements can be made to the algorithm or to assumptions or predictions that were made.

Advanced Persistent Threats (APT) are an especially big problem for enterprises.  These attacks are intelligently designed by teams of attackers and are highly targeted.  They utilize some of the latest technology and are usually based on extensive information gathered about the target from sources such as social media, the dark web, probes of public sources, dumps from previous hacks, and social engineering.

Once in place, APTs can operate covertly over an extended period of time, causing significant damage to the organization, its customers, services, and ability to do business.  Intelligent solutions are needed to combat these threats.  For example, Bitdefender’s machine learning system analyzes programs as they run to identify anomalous behavior.  It can identify potentially vulnerable software and alert administrators to this before those vulnerabilities are exploited by attackers.  This puts the enterprise on the proactive rather than the reactive side of security.

Machine learning systems need to be quite powerful so they utilize the power of the cloud to process large amounts of data and millions of distributed clients to collect it from around the globe.  Machine learning systems are comprised of multiple machine learning algorithms that each process the data in different ways, looking for patterns of attacks or anomalous behavior.  What once was science fiction is now science fact.

Such systems are proven technologies, not futuristic fantasies.  Bitdefender’s anti-exploit technology identified 100% of the Adobe Flash exploits of 2016 and an astounding 99.99% of malware.  Microsoft is using machine learning in their SmartScreen filter and Google uses it in their Safe Browsing initiative.  When tested against traditional security systems, machine learning systems resulted in fewer false positives as well as fewer false negatives, meaning that more attacks were thwarted and less time was wasted chasing false alerts.

For companies, this is a big saving to the bottom line and a cost-effective way to implement security.  Cybersecurity systems are more effective and keep their sensitive data away from prying eyes and key systems available for use while IT and security personnel are not distracted by as many false alarms so they can be focused on what matters, keeping the company safe.

Does your cybersecurity strategy include machine learning technologies?

About The Author

Eric Vanderburg

Eric Vanderburg is an author, thought leader, and consultant. He serves as the Vice President of Cybersecurity at TCDI and Vice Chairman of the board at TechMin. He is best known for his insight on cybersecurity, privacy, data protection, and storage. Eric is a continual learner who has earned over 40 technology and security certifications. He has a strong desire to share technology insights with the community. Eric is the author of several books and he frequently writes articles for magazines, journals, and other publications.

Leave a Reply