National Post (National Edition)
AI new tool in ‘cyber arms race’
MONTREAL • Advances in artificial intelligence are fuelling a new race between hackers and those toiling to protect networks.
Cybersecurity is always a race between offence and defence but new tools are giving companies that employ them a leg up on those trying to steal their data.
Whereas past responses to cybercrimes often looked for known hacking methods long after they occurred, AI techniques using machine learning scan huge volumes of data to detect patterns of abnormal behaviour that are imperceptible to humans.
Experts expect machines will become so sophisticated that they’ll develop answers to questions that humans won’t clearly understand.
David Decary-Hetu, assistant professor of criminology at the University of Montreal, says defenders have an edge right now in using artificial intelligence.
“But who knows what’s going to happen in a few years from now,” he said in an interview. “The main issue is that if you’re defending a system you have to be good 100 per cent of the time, but when you’re attacking the system you only have to be successful once to get in.”
Decary-Hetu said a growing list of corporate and government officials, including Bank of Canada Governor Stephen Poloz, say infiltrations are their top worry have a very good reason to fear.
The Bank of Canada warned in its semi-annual review this month that the high degree of interconnectedness among Canadian financial institutions means any successful cyberattack could spread widely throughout the financial system.
Reports suggest cybercrime costs the Canadian economy between $3 billion and $5 billion a year, including ransom paid to foreign criminals.
Hacks of Sony Pictures, Uber, Ashley Madison, Yahoo and multinational retailers have sparked unsettling headlines about security of personal information.
One of the latest to face scrutiny is global credit-reporting firm Equifax. Hackers accessed the personal information, including names, social insurance and credit card numbers, as well as usernames, passwords and secret question/secret answer data of 19,000 Canadians and 145.5 million Americans.
Niranjan Mayya of Toronto-based Rank Software, said it takes on average 143 days for a breach to be detected.
“Clearly the old style techniques of looking at cybersecurity threats and having people go through each threat aren’t working anymore, so automated means of detecting threats has become more and more important,” he said.
David Masson, Canadian manager for U.K.-based Darktrace, said AI will help to keep up with threats by quickly identifying and stopping attacks by picking up on subtle markers that identify bad behaviour.
“You’re kind of looking at a cyber arms race,” he said in an interview. “If you want to keep up with this threat and put the advantage back in the hands of the defenders you’re gonna have to use AI.”