Spotting the Links Between Fake Accounts to Root Out Fraud on LinkedIn, Facebook, and Twitter
In the wake of LinkedIn’s recent announcement regarding measures the organization is taking to rid the platform of fake accounts, it should first be said that the move deserves applause. Platforms suffer reputation damage that quickly translates to financial loss, and users suffer deception, deceit, and sabotage. This cannot be allowed if a social platform is to succeed and continue to provide its good users with a safe and trustworthy environment, and fraud of this kind must be stopped.
LinkedIn is the archetypical example of a platform that runs on trust — consider the stakes for a legitimate user of the platform. At a minimum, one is engaging in professional networking. More likely, people are actively seeking employment; often, under desperate pressure. Interactions on the platform can be literally life-changing; for better or for worse. One need only imagine the anguish of having been misled into thinking one had found a dream job, when in fact, it was a fraudster all along. This post from How to Geek details exactly how these kinds of scams transpire, and as reported in this news story from WTKR in Virginia, a woman not only didn’t get the job she thought she had; she lost thousands of dollars in the process.
LinkedIn isn’t the only social platform dealing with worrisome spikes in fake account fraud, as can be seen from the chart below:
What’s instructive about the LinkedIn story, is the proactivity they’re displaying in dealing with the issue. As noted in a blog post by LinkedIn’s Paul Rockwell, the company’s Head of Trust & Safety, LinkedIn’s efforts included, “Preventing 19.5 million fake accounts from being created at registration. This means the vast majority — 95% — were stopped automatically, without ever being live on LinkedIn.”
At DataVisor, we often speak of “account level detection,” and LinkedIn’s methods in this use case are a good representation of this kind of approach. By stopping fraud at the point of registration (i.e., at the account level), they are able to prevent damage before it happened. This isn’t the case with reactive solutions that focus on the aftermath of malicious actions (i.e., at the transaction level). Waiting for an action to take place, assessing its legitimacy, and trying to constrain the damage if it isn’t, is a recipe for disaster as it inevitably involves “collateral damage” — good users suffering terrible experiences. Mr. Rockwell, at the close of his post, offers a beautiful enunciation of the close connection between fraud prevention and customer experience:
“When we stop fake accounts, we start more chances for economic opportunity.”
Today, we have the technology to push back the rising tide of fake account fraud. It takes coordination at a large scale to mount the kind of attacks these modern fraudsters are launching, and we now have the ability to spot the patterns and connections that reveal these coordinated actions. With the power of advanced unsupervised machine learning at our disposal, we no longer have to sit on the sidelines, learning from our mistakes, and trying better next time. As LinkedIn is doing, we can get in front of fraud in real time, and stop damage before it happens.