Global Swift banking network readies AI platform to combat cyber threats – Source: www.csoonline.com

The global Swift (Society for Worldwide Interbank Financial Telecommunication) messaging platform, fundamental to the movement of money around the globe by institutions, is about to deploy its first artificial intelligence (AI) system to detect fraud in the vast flow of transaction requests that cross its network each day, the organization has announced.

The project to add AI fraud detection to the financial sector’s critical messaging platform was first announced as an experimental project in 2022.

More than two years on, and over a year behind schedule, Swift said it now plans to move from the AI trial phase to live fraud detection in January 2025.

The new capability will add anomaly detection to Swift’s existing Payment Controls Service (PCS) by screening payment instructions for missing data, as well spotting anomalies by comparing pseudonymized historical transactions with new flows.

Separately, Swift has worked with a clutch of leading financial institutions, including BNY Mellon, Deutsche Bank, DNB, HSBC, Intesa Sanpaolo, and Standard Bank, to test how AI can be used to share fraud data collaboratively using federated learning without compromising privacy or confidentiality.

A prime target

If a company or person wants to move money to an account in another country or institution, payment is initiated as a message from a bank through the Swift network. The role of Swift is to act as a secure channel through which communication about payments happens.

Inevitably, this has made the platform a prime target for hackers. In what remains one of the most notorious bank fraud incidents ever recorded, in 2016 North Korean hackers came close to stealing nearly $1 billion from Bangladesh Bank (a mistake by the attackers reduced the theft to a mere $81 million).

This, combined with smaller but still serious incidents, revealed the vulnerability of Swift to a range of compromises, including insider attacks. The organization was later heavily criticized for failing to address the weaknesses at smaller banks using the Swift system, despite knowing they existed.

Anomaly detection, federated learning

The latest announcement doesn’t explain the new AI in detail, but it’s known from previous announcements that the anomaly detection is part of a partnership with Red Hat, C3.ai, and Kove.

C3.ai will provide the AI applications, Red Hat OpenShift the containerization, and Kove the software-defined memory technology that helps efficiently scale and manage AI’s considerable memory demands.

The challenge of using AI for anomaly detection in a platform such as Swift is that the technology can’t get in the way by, for example, slowing down transactions with false positives or rejections. That’s why it’s not surprising that the organization has taken its time with the proof of concept phase before going live.

Since the Bangladesh Bank attack, attackers have grown more sophisticated, simultaneously probing multiple institutions for weaknesses. Banks could share data on anomalies to reduce the risk, but are reluctant to do so because of regulatory and privacy worries. Federated learning allows AI systems to be trained on collective data in a way that keeps the underlying data anonymous.

Responsible AI

Nevertheless, getting the organization’s 11,500 members to feel at ease with all this can’t have been easy. That’s why Swift previously announced an AI governance framework that emphasizes explainability, accuracy, and auditability, in compliance with ISO 42001, the NIST AI Risk Management Framework, and the EU AI Act.

“AI has great potential to significantly reduce fraud in the financial industry. That’s an incredibly exciting prospect, but one that will require strong collaboration,” commented Swift’s chief innovation officer, Tom Zschach.

“Swift has a unique ability to bring financial organizations together to harness the benefits of AI in the interests of the industry, and we’re excited by the potential of both of these pilots to help further strengthen the cross-border payments ecosystem,” he said.