Artificial intelligence (AI) holds great potential for cybersecurity processes and management, yet it is also a tool used successfully by cyber criminals. Swinging the balance in favour of good was the focus of a panel discussion during the August 5th session of ITWC’s MapleSec Satellite series.
For panelist George Nastasi, Associate Partner Security Threat Management with IBM Canada, it all comes down to AI’s ability to automate massive amounts of work and reduce the effort required to mount a large scale cyber attack. Citing examples such as algorithm manipulation, social engineering, and phishing campaigns, he said criminals recognize the value of AI in streamlining their tactics and techniques and allowing their exploits to launch faster.
Panelist Adam Frank, CTO Security Intelligence – also with IBM Canada – weighed in with the point that criminals can actually use AI to harvest the information they’ve stolen and produce more skillfully targeted profiles for their campaigns. “What we see normally right now is that breaches come predominantly through email or phishing campaigns,” he explained. “Those campaigns can be much more believable when they’re targeted using AI.”
Next Steps in AI
Fortunately, according to Frank, AI is also a growing weapon in cyber defence. “We have another layer of AI which can help us basically mitigate or reduce threats and highlight what’s real,” he said. “The next step is automating the response, so depending on what type of incident we’re dealing with, there are certain situations where the AI could automatically turn off the network connection and disable the account.”
The Upside of AI
According to panelist Dr. Cindy Gordon, CEO of Salesforce Inc., the healthcare industry demonstrates some of the most exciting uses of AI, with significant advances made possible by real time monitoring and wearable computing. “I think all of these self-monitoring systems with AI are real breakthroughs,” she said, “and not just in terms of early detection, but also in other areas. At the same time, we have to recognize that we’ve got a global network of cyber criminals, and only 1% are actually caught and convicted.”
“What I can add here is that the power of AI prediction, especially as strategy prediction, is increasing significantly with the integration of game theory and AI,” said the panel’s fourth member, Dr. Ali Dehghantanha, CEO of CyberOrion and Canada Researcher, Cybersecurity & Threat Intelligence with the University of Guelph. “A lot of exciting development is happening in the prediction area.”
Room for Improvement
In response to a question about the downside of AI technology from panel moderator Rob Meikle, Principal Consultant for Meikle 360, Dehghantanha acknowledged that there are many weaknesses, including the difficulty in seeing how decisions are made. “We should be able to identify the sources of those decisions in order to justify them and encourage others to trust the AI,” he said.
In Gordon’s estimation, there is a real need for education and knowledge in this area because of the pervasiveness of solutions that are now using AI capabilities. “We had to train people to understand profit and loss and balance sheets,” she said. “I believe that we need the same kind of business knowledge in order to understand AI.”
Moving on to a discussion of the skills gap, Frank echoed the comments made by guests on a cybersecurity panel. “We hope to address the skills shortage in cybersecurity with AI, but that’s just punting it down the road a bit because there’s still a problem finding the necessary skills to deal with AI,” he said. “From a skills perspective, we need people who can assemble a dataset, detect the biases within it, handle that data appropriately, and then know how to apply it.”
Commenting on the areas in which we should be using AI right away, Nastasi said time is of the essence. “There are so many things you can incorporate, whether it be monitoring security, response, identity, or access management,” he said. “There are ways to integrate AI and automation to be able to get more efficient, to sharpen your work, and reduce the amount of human-based manual labour from hours to seconds.”
The panel concluded with a discussion of the future of AI, with predictions for everything from encrypted AI to robot police forces. “We’ll probably see some interesting advancements in the next five to ten years,” said Frank. “It’s going to be very exciting.”