Raising the Bar

Gaining acceptance is a means for preventing online fraud

• By Allison Armstrong
• Mar 01, 2009

With every successful breach of Internet security, we learn the same lesson: there is no silver-bullet solution to prevent online fraud. Attackers are relentless innovators, single-minded in the pursuit of a breach.

Security professionals have long understood that any single factor—like a password—can be spoofed. Even security tokens, considered by some to be 100 percent reliable, can in fact be passed around, phished, lost or stolen.

Risk-based authentication is widely adopted in the financial services industry and is now gaining acceptance as a means for preventing online fraud. Risk-based authentication addresses the increasing complexity of threats by applying multiple factors to score the likelihood that a user is who he or she claims to be online.

Risk-based authentication has typically focused on assessing location and device as proxies for the user. Previously, there were no direct means to confirm the user, so security vendors have invented substitute methods that take the IP location, device signature or certificates as a substitute.

The problem with proxies in risk-based authentication is that they aren’t the user and can be used to spoof the security system. Attempting to plug this hole leads to other issues, including that many riskbased systems tend to exhibit a high falsepositive rate and proxy-based authentication is vulnerable to collusion, social engineering and account sharing fraud.

The solution might be to add a solid, user-level form of authentication that doesn’t depend upon proxies but also doesn’t require the cost and cumbersome implementation of physical biometric solutions.

Authentication Shrinks the Attack
Each authentication layer—network, device and user—is strong in blocking certain attacks but less effective against others. Any single defense is vulnerable to a particular line of attack. The most effective security combines factors at all three layers to provide the most resistant posture.

The network layer uses login IP addresses and their geographical origination points as a proxy for users. However, network-layer authentication alone isn’t capable of verifying the identity of a user. Similarly, device signatures verify the hardware as a proxy for the user, matching a known device signature with the current attempt. But what if the user’s laptop has been compromised, along with cached passwords?

The network and device authentication layers have limitations in blocking collusion, social engineering or accountsharing fraud.

New Frauds, New Responses
There are attacks that cannot be prevented by the network and device layers. As long as they are viable avenues for fraudsters, the attack surface remains a broad and inviting target.

Collusion. When willingly surrendered to a collaborator, credentials are a particularly damaging weapon against the systems they were meant to protect. Collusion destroys the very revenue stream of software and information service providers, when legitimate users invite others to use the service by riding in on their username and password.

Corporate social engineering. The same goes for the betrayal of trust inherent when a disgruntled executive assistant uses the boss’ own access codes to foment mischief. A coworker grabs a token and is able to compromise critical controls.

Friends and family social engineering. A child uses a parent’s work account or a friend steals a bingo card, token or security device. In one real-life example, the son of a real estate agent used his father’s password to log into privileged information on the MLS. The teen could discern which homes were vacant, and threw large, destructive parties in the homes.

How does an organization raise the bar to mitigate these threats? The answer is to integrate the network and device authentication factors with a user-centric layer, reducing the attack surface. In addition to authenticating at the device and network layer, AdmitOne Security Suite employs a user layer capable of detecting a user’s unique “typing signature.” Using keystroke dynamics, the user layer is able to mitigate threats from collusion, account sharing and social engineering and is able to work in concert with other factors to reduce the threat from additional attack vectors.

Keystroke Dynamics
While typing a string of characters in a password for example, every person exhibits a series of timing events—a typing signature. The measurement and comparison of user-specific typing rhythms is called keystroke dynamics.

The raw measurements of a user’s typing rhythm can be recorded from almost any keyboard to determine related metrics like dwell time and flight time. These and other timing measures form a series of mathematical data representing a user’s typing signature or template. AdmitOne’s patent-pending algorithms compare this data to previous attempts and calculate a confidence score that the user is who he or she claims to be.

Keystroke dynamics cannot be lost or easily stolen. It is portable and permanent with users, and cannot be carelessly “loaned” to another. This proven technology is recognized by analyst organizations as an important addition to enterprise security systems.

This article originally appeared in the March 2009 issue of Security Today.