How Truecaller is combatting the rise in deepfake fraud

Incidences of deepfake fraud have been spiking recently.

Feb 8, 2024 - 04:30

0 31

How Truecaller is combatting the rise in deepfake fraud

Spurred to new heights by generative artificial intelligence models, incidents of deepfake fraud and "identity hijacking" have surged in the past year, marked by a number of well-publicized instances involving President Joe Biden, Taylor Swift and multi-million-dollar scams.

Scammers, fraudsters and bad actors have taken advantage of cheap, highly accessible generative AI models in an attempt to steal money from corporations, publicly humiliate and harass certain individuals and affect election results.

Verification platform Sumsub released research in November that highlighted a "10x increase in the number of deepfakes detected globally across all industries" from 2022 to 2023, noting a 1,740% surge in fraudulent deepfake incidents in North America alone.

"The rise of artificial intelligence is reshaping how fraud is perpetrated and prevented," Pavel Goldman-Kalaydin, head of AI at Sumsub, said in a statement at the time. "Deepfakes pave the way for identity theft, scams and misinformation campaigns on an unprecedented scale."

The threat

Last year, a mother received a phone call from scammers who, claiming they kidnapped her daughter, were demanding a $1 million ransom payment. She heard her daughter pleading for help on the other end of the call and started to panic.

But her daughter's voice was an AI-generated synthesis; her daughter was at home, in bed. Safe.

In the lead-up to the New Hampshire presidential primary, a deepfaked robocall of Biden circulated, telling voters not to participate in the primary. New Hampshire Attorney General John Formella said the robocalls numbered in the "thousands."

"With the help of (Large Language Models) ... phishing and online fraud can be created faster, much more authentically and at significantly increased scale," a March 2023 Europol report reads.

More recently, a worker at a multinational company in Hong Kong was duped into giving scammers — who posed as convincing (deepfaked) versions of the company's senior officers in a video call — $25 million.

More deep dives on AI:

"If any of us think we're too smart to get scammed, we're not," Clayton LiaBraaten, the senior strategic advisor for Truecaller, an enhanced caller-ID app, told TheStreet.

The threat of AI-powered fraud is multi-fold, LiaBraaten said.

It goes beyond the speed and believability provided by the technology and into social psychology; where robocalling used to involve mass, broad campaigns, AI fraud can be much more targeted.

With AI, "it doesn't take a lot of machine learning-based data mining" to know where someone went to school, what kind of car they drive or how many kids they have," he said.

AI, LiaBraaten said, enables highly relevant, contextualized, "surgical" attacks on segmented iterations of vulnerable groups of people.

"The scammers and the fraudsters are very astute technologists and they're very capable social engineers," LiaBraaten said. "So now I've got this tech that makes it easier. I've got the psychology that's there. And so you think about all of the information we all put on social media ... there is currently no 100% solution."

"There are just better than nothing solutions and much better than nothing solutions," he added.

The 'much better than nothing' solution

When Truecaller launched in 2009, the idea was a simple one of providing advanced caller ID and spam blocking. The company, which has more than 360 million active users, has since diversified its offerings to include fraud detection and prevention.

Its anti-fraud solution currently involves a three-pronged approach.

The first involves number intelligence within Truecaller's ecosystem, where Truecaller uses machine learning to provide dynamic caller ID.

Truecaller looks at number behaviors when a call comes in — suspicious numbers stand out "because we've seen hundreds or thousands of calls that last two seconds."

Truecaller's community of users additionally allows for the flagging of known scam phone numbers.

"We get better at detection. They get better at avoiding detection." — Clayton LiaBraaten, senior strategic advisor for Truecaller

The company is also currently developing technology designed to detect synthesized voices. Part of this involves using AI to compare a given call to a database of voice prints of public figures, and the other involves the ability to identify certain artifacts in a synthetic voice —indistinguishable to the human ear — that are distinct from the digital voice of a human.

That method of synthetic voice detection, LiaBraaten said, will be available "within the election cycle."

But even in bringing this solution to market, Truecaller won't be able to escape the game of Whack-a-Mole that has become so common in cybersecurity: "We get better at detection," LiaBraaten said. "They get better at avoiding detection."

In its effort to protect its customers' phones, Truecaller has released a number of additional tools, including a verification and authentication process for official business numbers and call screening, another shield against potential scam calls.

"Can you imagine walking out of a Best Buy with a new laptop and not having McAfee or Norton on it? That's where we need to be," LiaBraaten said. "What we're trying to do is to put a lot more specificity and science around it and emphasize taking the right call, taking the call that matters. If you ignore calls, you're ignoring opportunity."

"Call screening, anti-fraud detection is the only way that people are going to restore their trust in communications and reclaim their phone."

Contact Ian with AI stories via email, [email protected], or Signal 732-804-1223.