Catching connected criminals: Exposing the networks driving modern fraud
Financial crime is now a complex web spun by coordinated networks—and investigators face a difficult task untangling it.
Financial crime is not a series of isolated events committed by individual actors.
Modern fraudsters coordinate, distribute and engineer their operations across fragmented systems, using interconnected mule accounts, synthetic identities, layered payment flows, and devices.
Each of these elements appears low-risk in isolation, but forms part of a system deliberately engineered to evade traditional monitoring.
This means the impact on financial systems can only be understood by examining interconnected actions that become visible when viewed together.
In this environment, the challenge for banks is no longer simply identifying suspicious transactions, but understanding how seemingly disconnected signals form coordinated networks.
The real threat facing financial services
In the UK, fraud now accounts for more than 40% of all crime, making it the single largest category of criminal activity. In the first half of 2025 alone, £629.3 million was lost to payment fraud and scams, a 3% rise on the previous year. These figures highlight the scale of modern fraud.
Spotting fraud is no longer about identifying a single suspicious transaction or isolated account anomaly. Increasingly, this type of crime manifests through coordinated, multi-layered activity. Mule accounts are linked to synthetic identities, compromised credentials are reused across platforms, and shared devices connect profiles that appear unrelated.
Funds are deliberately routed through layered transaction chains so that no single step appears unusual. Criminal groups fragment their operations across institutions, products, and jurisdictions, exploiting the seams between banking, payments, and insurance systems.
The challenge is intensified by AI, which is rapidly increasing the scale, speed, and sophistication of fraud. More than half of fraud cases now involve some form of AI-enabled tactics, such as deepfakes, synthetic identity generation, and AI-powered phishing scams. AI is enhancing traditional fraud techniques, increasing their scale, credibility and speed.
Against this backdrop, legacy fraud systems that rely on traditional databases to analyse account data in isolation are increasingly outmatched. While designed with rule-based controls to detect individual anomalies, they cannot interpret fluid, AI-enhanced networks that are constantly advancing and adapting. This makes graph data models imperative for banks that want to protect themselves against fraud.
READ MORE: Bank of England: Cloud outages and DDoS attacks pose risk of cascading systemic failure
From isolated alerts to connected intelligence
By mapping data relationships, graph intelligence allows banks and financial institutions to better understand how accounts, transactions, devices, and identities interconnect. As a result, it exposes clusters and hidden linkages that would otherwise remain buried across siloed systems.
By unifying customer, account, and device data as part of a graph-based detection model, institutions such as BNP Paribas Personal Finance have reported a 20% reduction in fraud losses and significant time savings per investigation.
For financial services, the value of graph intelligence extends beyond fraud prevention. By modelling the connections between customers, accounts, transactions, controls, and regulatory obligations, institutions gain a richer understanding of operational and compliance exposure.
Traditional systems assess regulatory impact in silos, requiring manual reconciliation across teams and datasets. Graph-based models allow institutions to trace relationships and dependencies, enabling faster impact assessments.
Making sense of networked fraud
As AI adoption accelerates, the role of relationship-rich data becomes even more significant. AI models perform best when they evaluate not just isolated data points, but the relationships between customers, accounts, transactions, and behaviours over time.
Models trained on isolated datasets may identify patterns, but struggle to interpret the broader network of dependencies that shape real-world outcomes. In contrast, graph databases introduce that contextual layer, enabling AI applications to operate with greater precision and governance.
To counter a networked threat, banks must move beyond isolated alerts and evolve their data strategies accordingly. Graph data models provide the structural visibility to connect accounts, identities, and transactions, enabling institutions to detect coordinated patterns faster and respond with greater precision.
When fraud is built on connections, understanding those connections is the ultimate defence.
Michael Down is Global Head of Financial Services at Neo4j
