If you’ve been following any of the recent PCI (Payment Card Industry) breaches, you’ll see two trends coming from the breached organization. One one hand, they’ll say that the PCI standard is flawed since it they were compliant and still got hacked. On the other hand, they’ll say the problem was their QSA (Qualified Security Assessor) messed up. Although both the standard could be better and a QSA can be more thorough, it’s time for organizations to stop passing the buck and admit that they screwed up.
Now I understand that there are certainly differences of quality in each QSA. I’ve seen everything from rubberstamped reports that make you wonder if the QSA can spell ‘PCI’ to a tinfoil hat QSA who uses the standard to create new, crazyrequirements. But in the end, I sometimes wonder if the organizations understand truly what makes a good QSA and what their role is. Ultimately, the QSA is supposed to be an auditor (yes, I know technically we are ‘assessors’). That means the QSA is trying to make sure the processes the organization established are working properly. In order to do that, the QSA needs to typically sample the controls and do some digging around. This should allow the QSA to determine that the control is generally operating correctly. However, the QSA is NOT the master of the organizatio’s destiny for security. Heck, the audit is only once a year and cannot turn over every stone. The organization owns and is responsible for the processes, the controls, monitoring, and reacting. If they break down, that is the organizatio’s problem.
Now as far as the standard goes, I hear a lot of people (generally non-QSA) stating that the PCI DSS is flawed. First of all, let’s just understand that this was written to push compliance, not security, to a certain level since it was severely lacking in most organizations. Just to be clear, compliance is just a hurdle and not a ceiling. What I do’t understand is then what other framework or compliance is better than PCI? For example, ISO 27001 is still too general and based on business decisions. The standard is still the most thorough, prescriptive, and layered one out there. This is not only for the controls within it but also the testing. For example, testing applies both externally and internally and escalates from scans, to pentests, and even web app reviews. For all the naysayers out there, I’d love to see a better version from them. Remember, PCI was the first and still the onyl one out there that has pushed hard for web application security such as the OWASP Top 10. Now for those who think the standard is’t detailed enough, give me a break. It is’t going to tell you exactly how to do X with product Y in a Z architecture. There are just too many variables.
So let’s look at the latest details on the big breaches. Now we are finally starting to get enough details to understand what happened. Let’s go with the assumption that the downfall to Heartland/Hannaford was SQL injection and see where the problem hypothetically is. Well the standard clearly states that it needs to be properly coded in requirements 6.5.2. Keep in mind that a QSA will not and may not even be able to review all your code. Oh yeah, that is the organizatio’s requirement in 6.3.7. This also needs further tested through a web application review (possible third party problem) or prevented through a web application firewall (organizatio’s problem) in 6.6. Not to mention that this would or should be detected during the vulnerability scanning (ASV problem) and penetration testing (probably third party problem) from requirements 11.2 and 11.3. So how exactly did the standard fail here when there are at so many layers of controls and testing designed to stop this?
This brings me back to yet another ranting point. It is unbelievable the number of organizations who are constantly trying to find the absolute cheapest vendor for any of the stuff we just discussed including the QSA, the ASV, and other testing. There is little going into understanding the quality or qualifications of the vendors being selected by the organization. This absolutely leads to ‘getting what you pay’ for quite a bit. But just remember, that is a choice of the organization to pick a deficient vendor.
We also need to realize that even if SQL injection had worked, there were still plenty of other controls within PCI that could have stopped the breach. First of all, there is a anti-malware requirement. Yes, I know that only recently was it clearly required for all platforms, not just those ‘commonly affected’. But we have seen Linux malware and it’s supported by all commercial anti-virus vendors and even free ones. So did the organization just not do it because it was’t clearly dictated as a hurdle within the standard? I guess so. There is also a requirement to restrict firewall traffic both inbound AND outbound. So why was the malware allowed to talk to the internet from payment systems? There is no good business reason there. So the standard certainly is’t as flawed as people want to assert, though there it could be improved.
In the end, we certainly can conclude a few things. First of all, despite the fact that these organizations had been assessed as compliant, the organizations clearly were not compliant. Though it is possible that a QSA missed something, that is still the organizatio’s problem. Also, the standard cannot be perfect but it is very well built and, if followed, would have stopped these. In security, like any process, it may only take one flaw for a breach to occur – even if that was the standard or the QSA. But that’s why security has to be layered and has to be based on quality, not compliance. Regardless, the whole security process is owned by the organization and they need to take responsibility when it breaks. Passing the buck does’t undo the breach, but does make for some interesting blamestorming to watch.