Kantardzic explores the future of Machine Learning

June 1, 2018

Mehmed KantardzicIn the May issue of Ubiquity, an Association for Computing Machinery publication, Dr. Mehmed Kantardzic and PhD student Tegjyot Singh Sethi contributed the article “Big data: when good machine learning leads to bad security.” An informed opinion piece, the paper posits that as it relates to security, machine learning is only useful as a dynamic, rather than static solution.

“All of these machine learning techniques, we are assuming we’re building in an ideal world," said Kantardzic. "We are assuming we are building the best techniques, that good people, bad people, they are always staying the same. We are assuming that they will always stay the same.”  

Examples of machine learning in a security application includes captcha systems, spam filters in emails, and click through fraud protection. While employing a software package to resolve those issues may prove efficacious initially, Kantardzic believes that attacks to the system will eventually evolve to circumvent those solutions.

“There is no solution that is 100%,” Kantardzic said. “It is some kind of game between developers and attackers. It’s always some kind of fight which is infinite fight and infinite solution at the beginning as soon as possible.”

One possible resolution to this dynamic problem, is to employ a series of watcher’s on the wall to check and recheck the security of the program. Furthermore, Kantardzic hopes that in employing data mining and machine learning techniques, that a system can be developed to monitor and anticipate new attacks as they are developed. In doing so, security solutions can expand beyond just prevention to detection, creating a cohesive whole to an ever evolving problem.

“The basic idea is there is no finite solution for machine learning, especially in cyber security,” said Kantardzic. “It is necessary for people to come together to make this comprehensive solution, not as a software package, but to give real solutions in these domains.”