Botnets, with odd names like Mirai and Dorkbot, give bots a bad name. That being said, it’s important to understand that just because something is called a bot doesn’t mean it’s bad. We use the term to refer to almost anything from the spider programs that search engines like Google use to make a searchable index of your site to the most malicious tools hackers use to steal data. The reality is that most tools we call bots are beneficial to the workings of your site and the Internet. Intent and usage make a huge difference, even for the most helpful bots.
IT and security teams must be prepared to manage bots with their eyes open. According to Akamai’s Fourth Quarter, 2017 State of the Internet/Security Report, bot traffic accounts for more than 30% of all pure Web traffic (excluding video streaming). Web teams can’t afford to ignore the impact that bots have on their systems, nor can they block bots entirely.
There are also some bots that absolutely, positively must be defended against. Credential abuse bots, those that check an email and password against your site login page, are responsible for 43% of all login attempts, according to the Akamai data. There are literally billions of fake login attempts happening every month. And if you’re running a hotel or airline website, a significant majority (83%!) of all logins against your site are driven by malicious bots.
Bots and botnets are all too easy to ignore for many organizations. They’re part of constant noise that every site sees, something that quickly gets filtered out of our ability to care about. But that’s part of what makes bot traffic so dangerous. Credential abuse attacks have been around since shortly after the first Telnet and Secure Shell servers were exposed to the Internet, and scanners looking for services will probably be some of the last systems to go offline decades from now.
The modern incarnation of account checkers and credential abuse bots have evolved significantly since those early days. Even a few years ago, one of the common ways to detect and protect against these tools was to use rate limits. If you saw dozens or hundreds of login attempts from a single host, you could block that host and move on. In response, bot developers have moved from single/few host models to using whole botnets to host their tools. They circumvent rate limits by using a few nodes of the botnet at a time against a single target, cycling through IP addresses so a single site only sees an IP once in a great while.
A few years ago, it might have taken half an hour for a system exposed to the Internet to see the first login attempts show up in the logs. In today’s world, it’s often just seconds after the system goes live that the first attacks happen. It’s not because attackers are so great that they detect your systems when they first come online. Instead, it’s because there’s so much bot traffic — acting as a kind of background radiation bouncing around without a specific target — that any system is going to be hit by random attacks from the very start.
Solving the Problem
Organizations looking to solve this problem can look to the efforts in fraud prevention to solve the problem of distributed credential abuse by these botnets. For instance, credit card companies have to look across multiple organizations and look for out-of-the-ordinary behavior. It sounds easy when stated like this, but the reality is that it requires visibility into traffic at a global scale and can’t be done by looking at the traffic of just one organization.
If you’ve ever received a call from your credit card company telling you that it put a hold on your account because it saw a charge coming from Ukraine just minutes after you purchased gas in Boston, you have an idea how this works. By themselves, neither the gas purchase nor the purchase from Eastern Europe are necessarily suspicious. But when intelligence from the two events is combined, a clear picture showing the impossibility of traveling from one location to the other is revealed.
Credential abuse bots can be detected in a similar manner. Traffic from an IP that hits your site once then moves on might not trigger any defenses, because a single event isn’t enough to draw conclusions. But if you can combine information from multiple organizations and see a pattern of that IP hitting a series of sites, a clear pattern of abusive behavior can be determined. In turn, this allows for signaling to other organizations, letting them deny the attacking IP automatically.
Bad guys are getting smarter and distributing their attacks. In return, defenders have to maintain controls that rely on what similar organizations are seeing. An attack that looks like it’s just part of the background radiation of the Internet takes on an entirely different meaning when you can draw patterns from shared information. Account takeover bots try to hide their activity in the noise of the bots that you want to have access to your site, which means you need to cut through the noise to find them.
Martin McKeay is a Senior Security Advocate at Akamai, having joined the company in 2011. As a member of Akamai’s Security Intelligence Team, he is responsible for researching security threats, customer education, and industry intelligence. With more than 15 years … View Full Bio