Tongdun Technology using AI to help firms fight internet fraudsters

Since the first World Internet Conference was held in Wuzhen water town in 2014, Zhejiang province has spared no efforts in developing its digital economy centered on the internet. For instance, in the first three quarters of 2018, the province's value added in major digital industries reached 385.6 billion yuan ($55.71 billion), up 14.8 percent year-on-year.

As a Fintech startup participating in the Wuzhen conference for the third time, Tongdun Technology has experienced robust growth in step with the WIC, according to its CEO and co-founder Jiang Tao. A Hangzhou-based service provider of intelligent risk management, Tongdun Technology uses artificial intelligence (AI) and big-data analytics to help financial institutions and internet companies combat internet fraud.

"Thanks to the internet conference, Zhejiang has become an increasingly appealing location for high-tech companies and startups," Jiang said. "In my opinion, WIC is gradually shifting its focus to AI and big data, which proves we have chosen a right market."

Tongdun will showcase its achievements in anti-fraud and overdue payment collection technologies at the ongoing WIC in Wuzhen, including the Overdue Payment Elf, which is an intelligent chat bot that can make phone calls to "delinquent" customers, even guide them through the payment process.

"Our collector bot already can ... interact with (human beings)," Jiang said. "It's very likely that in the next three to five years, when a bank calls you up for a late payment, it would not be a human being on the other side of the line but a bot, although you can't tell much difference between them."

Early results from a few trials indicate the streamlined and standardized chat bot, which is always on call and free of human errors, works particularly well in the early stage of "delinquency", when the due balance is still within or not far beyond the grace period.

A frontrunner in intelligent risk management, the unicorn startup has greatly reduced the risks and losses caused by account, card information and identity thefts. An anti-fraud innovation Tongdun is showcasing this time is voiceprint recognition. The technology seeks to identify individuals by analyzing their vocal characteristics. A fraudster's tone might not sound so different to your ears but the subtle difference in the voice's vibrations might be a useful clue for Tongdun's algorithm.

During a recent visit to Tongdun's office-cum-lab, a big screen on a wall was flashing rapidly, changing data and graphics. That was Tongdun's real-time monitoring panel, showing among other things the day's "risk events". The number hit 23,307,360 and was still increasing.

Tongdun's integrated intelligent risk management system processes Application Programming Interface calls made by customers each day. The system contains a database of 1 million online potential fraud suspects. "Traditional financial institutions, including banks, insurance companies and consumer finance companies, tend to lag behind in risk management," Jiang said, adding that he believes the biggest internet fraud risk comes from account theft.

Tongdun statistics show abnormal logins and fake phone numbers pose the biggest danger, accounting for about 48 percent of all risks. Other significant red flags include fraudulent activation and password cracking, which make up 15 percent and 12 percent of the risks, respectively.

"We prevent such crimes by analyzing accounts and transactions with algorithms and models," Jiang said. "Our technology can detect whether or not the real account holder has logged in, by analyzing all kinds of structured data." Even a user's mouse-clicking or typing pattern can be useful data. Once Tongdun detects a login is likely a fraud, it will instantly warn - before the fraudster can reach the next step - the financial institution allowing it to activate manual interventions in the following operations, such as a phone call to the account owner before a transaction is processed.

"In terms of accounts and transactions, we can reduce the rate of fraud to below ten-thousandths, or even one-hundred-thousandths, compared with between a thousandth to five thousandths now," Jiang said, adding that he believes increasing the perpetrator's costs is key to combating fraud. "When the costs become very high, they might quit ..."

Jiang, who once worked for Alibaba as technical director of risk management and anti-fraud services, will deliver a speech on AI application in the financial area on the sidelines of the conference.

Founded in 2013, Tongdun has served more than 10,000 customers in sectors such as peer-to-peer lending, microfinance, funds management and third-party payment. And so far, it has reportedly raised about $200 million of financing.