On the economics of ransomware

We blinked, and the world changed on us.

This [long] post is not meant as doom&gloom on the scourge of ransomware, but rather a look at some basic economic aspects of this type of attack, and some recommendations for the future.

So far,  2016 is definitely the year of ransomware. Every vendor is talking about it in their pitches, the media is all over it (good articles here and here), etc. This blog just adds to that cacophony, though hopefully adding a different perspective.

“Prior Art”: Lots of people are now talking about ransomware, and I’m sure many have in the past too. I’d be remiss if I didn’t point out that Anup Ghosh of Invincea wrote a scaringly prophetic blog post on this back in July of 2014! Check it out here. Also, I liked Daniel Miessler’s piece here.

Note : as I discuss these topics, I may sound insensitive to the plight of the victims. It’s absolutely not that: I think ransomware is a scourge that should be eradicated, that we bring to bear the full force of law enforcement, but I’m pessimistic it can be done.

There are several aspects of ransomware that make it interesting from an economic angle. Let’s explore some of them.

The “Taming” of Externality

First and foremost, to me, ransomware is the first major, widescale threat that significantly reduces the inherent externality of poor security practices. What does that mean?

Up until now, poor security practices by end users resulted in relatively light consequences for the users themselves. In some cases, being used as a spam relay might have been not noticeable, or at worst there was a rare circumstance where malware resulted in having to reformat one’s PC. Yes, annoying and potentially painful, but manageable. From a behavioral economics perspective, biases such as mental accounting made it even less painful.

The broader costs of that infection – spam being generated, systems that had to be wiped, etc… – were largely invisible to the user. In market economics terms, all these costs were externalities. This means that the agent in the transaction – the user – was not taking those costs into consideration when making their choice – in this case, poor security practices that let to an infection.

Enter ransomware. Now, the user is faced with the painful choice of paying the ransom – actual monies being stolen – or facing the imminent destruction of their data. Worse, depending on how that strain of ransomware behaves, it infected network drives and potentially backups as well. This triggers several well-known behavioural quirks/biases, including:
  • The salience of paying. It’s pretty clear that there is money being lost, and it’s your money (or your organization’s).
  • The immediacy of the request. It’s not something that can be postponed. Criminals know this, and exploit it: in many cases, ransoms increase as time passes.
  • Loss aversion. From Kahneman and Tversky’s work, we know the tendency of people to be loss averse.

All of this is, naturally, horrible for the user. From an economic perspective, though, it is interesting that this, in a way, “reduces” the externality of a poor security choice. The user now knows full well that their poor choice/practice may result in a non-negligible cost. [Edit: as someone provided feedback to me, just another way of saying “the chickens come home to roost”.] They’re understandably concerned, and rightly so. I don’t see this diminishing soon.

“To Pay or Not To Pay, that is the Question”

The second interesting point is analyzing the dilemma of deciding to pay the ransom or not. Even law enforcement seems ambivalent: recent advice has included both “pay” and “don’t pay”.

There’s two things to look at:
  • First, from a societal perspective, the issue is similar to the Tragedy of the Commons, a well-known economic problem. In the traditional Tragedy of the Commons, individuals overconsume a shared resource, leading to depletion. In the case of ransomware, it’s not the same: to me, it is close to the “Unscrupulous Diner Dilemma”, a variation of the more traditional Prisoner’s Dilemma, but where a group ends up paying more, even though they all wished they couldn’t. In the case of ransomware, the individual decision to pay negatively affects the community by supplying the criminals with additional funds to reward them for the crime, along with funds to reinvest in future capabilities for the tools, thus costing more in the future.
  • Individually, people and organizations should recognize that the rational economic decision is not just simply “is the cost of paying the ransom less than the loss associated with losing the data”. The decision should be based on that cost, sure, but also taking into account:
    • Is that the end of it? Will paying the ransom this one time be an exception? In most cases, hardly… As ransomware proliferates, different gangs will keep attacking.
    • Will paying the ransom even get the data back in the first place? As @selenakyle nicely pointed out recently, there’s little recourse if things goes wrong…

At the end of the day, we’re back to externalities:
  • Those recommending “don’t pay” don’t bear the cost of the advice: lost data, etc…
  • Those choosing to “pay” don’t [immediately] bear the indirect cost of enabling the criminals to continue their efforts.

A more realistic approach to handling of ransomware should keep these in mind.

“Thy ransomware’s not accidental, but a trade.”

There seems to be consensus that what has enabled the rise of ransomware is, among other things, the maturity of bitcoin. That was the point clearly made by @jeremiahg and @DanielMiessier (here). I agree: bitcoin seems to have tipped it, but along with other changes to the overall ecosystem that appear to have made ransomware a more viable attack.

Like legitimate business, criminals have explored ‘efficiencies’ in their supply chain. As the main example: bitcoin (the peer-to-peer exchange, not the currency itself) has removed significant “friction” from the system. Whereas before, the steps needed in the cashout scheme might include several steps – all of which incurred fees to the criminal – the ubiquity of bitcoin has made the cashout process faster, cheaper. Taking out the middlemen, if you will.

Regarding bitcoin specifically, there’s a couple of interesting points:
  • more than anonymous “enough”, bitcoin is a reliable and fast payment system. Even though it doesn’t provide full anonymity – the transactions on the blockchain can be traced to wallets – bitcoin is sufficiently opaque that the tradeoff of limited tracking with ubiquity/speed made it the currency/payment system of choice.
  • This leaves an interesting question about the bitcoin exchanges: Can we expect the exchanges to work against their own self-interest in restricting these transactions? What sort of defensive approach can we expect the exchanges to take? The danger of people equating bitcoin with ransomware is real, and the industry is right in defending itself.

All in all, from looking at the underground ecosystem, it looks like ransomware is a ‘killer app’: profitable, easy to use, etc…

“Much Ado About Nothing”? Maybe…

Finally, ransomware seems to have exploded into our collective attention, but is it really such an epidemic? While we deal with the onslaught of news/articles/posts about ransomware (including, of course, this post …), let’s recognize that there is vey little incentive to “underreport” ransomware infections. To wit:
  • InfoSec vendors can point to ransomware as the new ‘boogeyman’ that every organization should spend more resources to protect against.
  • Internally within organizations, like with “Y2K”, “SOX”, and “PCI” before it, we can now possibly start to see “ransomware” as the shibboleth that enables projects to be funded.
  • Media sites latch on to the stories, knowing the topic draws attention. As an example, a lot has been made of the incident where a Canadian university opted to pay $20,000 CAD . Would there have been the bombastic coverage if the cause of the loss was, say, a ruptured water main caused by operator error? Not likely…

I can’t help but wonder if this is not a manifestation of a couple of things:
  • one, a variation of what’s called the Principal-Agent problem: in an economic transaction where there is an expectation that an agent will act on behalf of a principal, but instead acts on their own benefit. In this case, bolstering the issue of ransomware above and beyond other relevant topics.
  • two, just your garden variety ‘availability bias’ from behaviour economics, where the ease with which we recall something inflates its perceived rate of occurence.

In either case, we can take a peek at the well-known Verizon Data Breach Report. What do we see? Verizon’s DBIR shows that ransomware, even as a subset of crimeware, is not as prevalent as other attacks. See figure 21 on page 24 of the 2016 report.

“’Advice’ once more, dear friends, advice once more”

Wrapping up, then. There is a fantastic paper by Cormac Herley, from Microsoft Research – So Long, and No Thanks for the Externalities – in which he discusses how users ignoring security advice can be the rational economic decision, when taking into account the costs of acting on some security advice. The paper is from 2009 and is still extremely relevant. I consider it mandatory reading for any security professional.

Taking that into account, how should we frame security advice about ransomware?
(One could argue whether ransomware is not exactly the change in cost that invalidates the conclusions. Might be an interesting avenue to pursue…)

At least to me, too much of the security advice we see about ransomware is not taking into account the aggregate cost of acting on such advice.

Ultimately, the protection methods have to be feasible to be implemented. With that in mind, here’s a few recommendations.

For individuals:
  • Be aware of your own limitations and biases as you interact online. To the extent that it is possible, incorporate safe practices.
  • Leverage the automated protections you have available – modern browsers have sophisticated security features, your email provider spends a ton of resources to identify malicious content, etc…
  • Devise and implement a backup system that fits your comfort level, balancing the frequency of backups with their associated hassle.
  • Periodically check and possibly reduce your exposure by moving content to off-line or read-only storage. Just like you wouldn’t walk around at night in a risky neighbourhood with your life savings in your pockets, make it a practice of limiting how much data is exposed.
  • If infected, don’t panic. Keep calm and, if you choose to do so, act promptly to avoid the increases in demands.

For corporate users, similar advice applies, boiling down to “don’t base your security architecture on the presumption that users are infallible at detecting and reacting to security threats”. Back it up with technology. On a tactical level, a few extra things come to mind:

  • Verify that current perimeter- and endpoint-based scanning of executables/attachments is able to identify/catch current strains of malware (ask your vendor, then check to make sure). It might be a sandbox approach, endpoint agents, gateway scanning, whatever. Belt & suspenders is a good approach, albeit costly.

  • Consider application-level monitoring for system calls on the endpoints. This includes watching for known extensions, as well as suspicious bulk changes to files.
  • Consider monitoring data-center activity for potential events of bulk file changes such as encryption. Yes, there may be false positives.

  • Re-visit the practices that allow users to mount multiple network shares.
  • Make sure the Incident Response playbooks include ransomware as a specific scenario. Prepare for single-machine infection, multiple machines hit, as well as scenario where both local and networked files are encrypted. While I’m skeptical of survey data such as this, getting familiar with how bitcoin transactions work might be a worthwhile investment.

To me, ransomware is here to stay: it leverages too many human and economic aspects to simply vanish. As with many other “security” issues, this is just another one that was never just a technology problem, but a social and economic one. InfoSec professionals should keep that in mind, remembering that the solutions are not always technical…

And to purveyors and enablers of ransomware, “a plague on your houses!