The rise of layered fraud prevention
By 2014, 15 percent of enterprises will adopt layered fraud prevention techniques for their internal systems to compensate for weaknesses inherent in using only authentication methods, according to Gartner.
Gartner analysts said no single layer of fraud prevention or authentication is enough to keep determined fraudsters out of enterprise systems. Multiple layers must be employed to defend against today’s attacks and those that have yet to appear.
“Malware-based attacks against bank customers and company employees are levying severe reputational and financial damage on their victims. They are fast becoming a prevalent tool for attacking customer and corporate accounts, and stealing sensitive information or funds,” said Avivah Litan, VP and distinguished analyst at Gartner. “Fighting these and future types of attacks requires a layered fraud prevention approach.”
Ms. Litan explained that while the layered approach to fraud prevention tries to keep the attackers from getting inside in the first place, it also assumes that they will make it in, and that multiple fraud prevention layers are needed to stop the damage once they do. She said that no authentication measure on its own, especially when communicating through a browser, is sufficient to counter today’s threats.
Gartner breaks down fraud prevention into five layers:
Layer 1
Layer 1 is endpoint-centric, and it involves technologies deployed in the context of users and the endpoints they use. Layer 1 technologies include secure browsing applications or hardware, as well as transaction-signing devices. Transaction-signing devices can be dedicated tokens, telephones, PCs and more.
Out-of-band or dedicated hardware-based transaction verification affords stronger security and a higher level of assurance than in-band processes do. The technologies in this layer can be typically deployed faster than those in subsequent layers and go a long way toward defeating malware-based attacks.
Layer 2
Layer 2 is navigation-centric; this monitors and analyzes session navigation behavior and compares it with navigation patterns that are expected on that site, or uses rules that identify abnormal and suspect navigation patterns. It’s useful for spotting individual suspect transactions as well as fraud rings. This layer can also generally be deployed faster than those in Layers 3, 4 and 5, and it can be effective in identifying and defeating malware-based attacks.
Layer 3
Layer 3 is user- and account-centric for a specific channel, such as online sales; it monitors and analyzes user or account behavior and associated transactions and identifies anomalous behavior, using rules or statistical models. It may also use continuously updated profiles of users and accounts, as well as peer groups for comparing transactions and identifying the suspect ones.
Layer 4
Layer 4 is user- and account-centric across multiple channels and products. As with Layer 3, it looks for suspect user or account behavior, but it also offers the benefit of looking across channels and products and correlating alerts and activities for each user, account or entity.
Layer 5
Layer 5 is entity link analysis. It enables the analysis of relationships among internal and/or external entities and their attributes (for example, users, accounts, account attributes, machines and machine attributes) to detect organized or collusive criminal activities or misuse.
Ms. Litan said that, depending on the size and complexity of the end-user institution, implementing the systems that support a layered fraud management framework can take at least three to five years, especially when it comes to the upper layers — Layers 3, 4 and 5. These efforts are continuous, because fraud prevention rules and models require ongoing maintenance, tuning and care.