Asian FIs turn to ML amid tighter rules, increasing fraud

Asia’s top financial institutions (FIs) continue to be challenged by the expanding threat landscape in the past few years as well as the tightening of regulatory and enforcement measures, according to a TABInsights study commissioned by GBG.

The survey was conducted across seven key Asian markets including China, Indonesia, Malaysia, The Philippines, Singapore, Thailand, and Vietnam with 250 respondents from the FI sector.

The study found more rigorous supervision and penalties have resulted in increased monetary losses in the form of regulatory fines, the highest-ranking component of fraud loss for 41% of FIs — a change from the previous survey in 2020 wherein direct fraud losses was ranked highest.

Also, the survey found that Asian FIs complete a higher number of transactions through mobile and online channels with the highest digital channel adoption seen in Indonesia (71%), closely followed by Malaysia (70%). 

Respondents said they expect the average daily digital transaction volume to surge by 70% in 2025 compared to 2022. 

As more FIs expand their digital offerings in response to consumer behaviour shifts towards mobile and digital, managing the cost of increasing compliance has emerged as a key concern for 70% of FIs, while the ability to scale fraud detection measures to growing digital transaction volumes (39%) and identity verification (33%) rank as the top challenges.

“Open banking, interconnected devices and ecosystems, and increased digital adoption in general has elevated the risk of digital fraud and cyber-attacks, and expanded the perimeter of attack FIs face today,” said Dev Dhiman, GBG’s managing director in the Asia-Pacific region.

“New technologies are being exploited every day by innovative perpetrators who continue to challenge FIs to escalate their technology risk management strategy and capabilities to comply with the increased scrutiny by regulators and customers alike,” said Dhiman.

The study revealed strong machine learning (ML) adoption in Indonesia (71%) and Thailand (69%), while third-party data is used more actively in China (77%), Vietnam (73%) and the Philippines (68%) alongside robotic analytics in Singapore (63%) and Malaysia (62%) to address false positives, indicating a maturing ML landscape among Asian FIs. 

While the region saw increased adoption of ML-based algorithm tools with automated smart models to address fraud prevention in the sector—with 47% of FIs actively using ML tools and 37% beginning to use them—one of the biggest challenges for these organisations undergoing digitalisation is the increased complexities in addressing data standardisation and governance to scale fraud detection.

About 38% of FIs indicated that inadequate data standardisation is their most critical gap, alongside 32% who are challenged by fragmented data because of piecemeal systems and software. 

In Thailand and China, fragmented data emerged as the top challenge for FIs. In Malaysia, after inadequate data standardisation, lack of good link analysis was also highlighted by 23% as the top challenge. Meanwhile, 59% of FIs said they increasingly rely on third-party data, alongside 58% that use ML to address false positives.

The evolving risk dynamics in the industry are forcing institutions towards stronger data integration and technology tools to future-proof their fraud risk capabilities and ability to gather effective data insights. Increasingly, institutions seek to integrate a spectrum of transactions, devices and big data to strengthen fraud detection capabilities. 

Data from interconnected devices are being used by 78% of FIs, while 76% use transaction data and 64% public data. 

In fact, 42% of FIs indicated their need to prioritise and invest in one platform to interexchange application data and transaction data in 2022. Within the next year, 47% of FIs plan to add internal unstructured data and geographic data to deepen their fraud detection.