Sami Khoury says there is currently no reliable way to “de-authenticate” malicious fake videos and that a voicemail message at the end of an ad “won’t do it anymore.”
Article content
OTTAWA — Canada’s cybersecurity czar says he doesn’t yet know how to publicly “de-authenticate” fake AI-powered videos, as his agency warns of the growing threat of disinformation campaigns in the upcoming election.
Anyone who has seen a political ad on television has probably heard it end with a line confirming that the message was paid for and supported by a specific party or candidate.
Advertisement 2
Article content
Article content
But what if a person in a video looks and sounds exactly like a politician or public figure and the message confirming that they approve of their content is completely fake and generated by artificial intelligence?
The possibility is real, it will likely happen sooner rather than later, and current verification tools are no longer enough, says Sami Khoury, director of the Canadian Center for Cyber Security (CCCS). A simple voice message at the end of an ad “won’t do the job anymore.”
“AI can now be used to almost impersonate my voice. This is the next evolution. Now you can take a clip of my voice, 30 seconds, a minute, and have it say something completely opposite of my message and it will be very authentic,” Khoury said in a wide-ranging interview with the National Post.
“This can be done quite easily using online tools. And then you evolve a little further and you get into deepfake videos,” he added. “Technology is moving in this direction. We don’t yet know how to authenticate… or deauthenticate. How can I tell it’s not my voice, or how can I authenticate that a message is truly coming from me? »
Article content
Advertisement 3
Article content
Khoury said part of the solution could be some form of “digital signature” placed on recordings that would help prove their authenticity, but the technology is not yet there.
“We know how to create digital signatures in emails, but then how do we now move them into the 2D world or the 3D world? he said.
The threat that AI-generated misinformation and disinformation pose to Canadians, particularly around election time, is not new, but experts agree it is increasing at a frightening rate.
In December Cyber threats to Canada’s democratic process report, the CCCS warned that “foreign adversaries” will “likely use” generative AI – tools that create synthetic images like DALL-E or Midjourney – within the next two years.
They will also likely create “deepfake” videos and images of politicians to influence Canada’s upcoming federal elections.
But the ability of opponents to generate fake videos and images “exceeds our ability to detect them,” the report notes.
“The airwaves can be polluted by fake automated calls, or fake videos… or fake messages or emails from people pretending to be the candidate,” Khoury illustrated.
Advertisement 4
Article content
Related Stories
These tools are also used to perfect a previously rudimentary email scam: phishing.
Where once a fraudulent phishing email was riddled with errors and contained ridiculous offers like fake diamond reserves in Africa for a “holding fee,” Khoury says the messages are becoming more sophisticated and dangerous.
“The days when a phishing email … contained typos, funny punctuation marks and sold you something too good to be true are long gone,” he said.
Now emails are almost perfectly worded and scammers use digital tools to learn more about the potential victim and craft emails tailored to their interests, he said. In some cases, you don’t even need to open the message for your computer to become infected and hackers to gain access to your system.
Khoury says hackers and hostile states are increasingly turning to phishing as it becomes more expensive and time-consuming to break into organizations’ systems from the outside as they strengthen their cyber defenses external.
Advertisement 5
Article content
“Companies are investing to make their products a little more secure. So the only way to get around that hard shell, that security perimeter, is… to catapult yourself into the middle of a network,” he said. “Phishing is usually one way to do that.”
Khoury says China, Russia, North Korea and Iran remain Canada’s primary cyber adversary states and the CCCS continues to see nation states attempting to breach government systems.
In June, Prime Minister Justin Trudeau’s national security adviser Jody Thomas also said for the first time that India was among the top sources of foreign interference in Canada.
The country’s relations with India were further strained two months later when Trudeau announced that Canada suspected India of being linked to the killing of a Sikh separatist activist in British Columbia.
Khoury declined to say whether India is now considered a key cyber threat actor against Canada, but noted that India is an “advanced nation” that has “cyber capabilities that we need to be aware of.”
Another growing cyber threat is ransomware, which essentially locks an individual or organization out of their network or data until they pay a significant fee.
Advertisement 6
Article content
According to a new survey by cybersecurity company Palo Alto Networks, the average ransom paid by Canadian organizations in 2023 was $1.13 million, double the amount reported in 2021.
In light of the growing ransomware scourge, Khoury says the Communications Security Establishment has quietly launched a program to warn other government agencies as well as external organizations of suspected imminent ransomware attacks.
Before a ransomware “payload” is delivered to a network, Khoury says there is a noticeable “dance” between a malicious actor and an organization’s infrastructure, with the former trying to break in the first time. as deep as possible in the second.
Khoury says that since the spring, his organization has privately issued about 500 “pre-ransomware” notifications.
“We’ve developed a technique to detect some of these dance moves with enough confidence now that we can issue an automated alert to say…we’ve picked up some of these signals, that there’s activity on your infrastructure. these are steps towards a potential ransomware incident,” he explained.
“In many cases, the feedback we hear is that it made a difference and that they were successful in isolating the system and preventing the deployment of ransomware,” he added.
Our website is the place for breaking news, exclusive scoops, long reads and provocative commentary. Please bookmark nationalpost.com and sign up for our daily newsletter, Posted, here.
Article content