The New PCI 6.6

All Your Public facing Web Apps Are Relevant To Us.

I’m going to start off this post with the moral of the story: Good intentions often have bad, unintended consequences. The following is the ‘Testing Procedures’ text of requirement 6.6 from the new PCI DSS v1.2 (source: [https://www.pcisecuritystandards.org/security_standards/pci_dss_download.html] (https://www.pcisecuritystandards.org/security_standards/pci_dss_download.html)):

For public-facing web applications, ensure that either one of the following methods are in place as follows:

• Verify that public-facing web applications are reviewed (using either manual or automated vulnerability security assessment tools or methods), as follows:

  • At least annually
  • After any changes
  • By an organization that specializes in application security
  • That all vulnerabilities are corrected
  • That the application is re-evaluated after the corrections

• Verify that a web-application firewall is in place in front of public-facing web applications to detect and prevent web-based attacks.

Note that for ease of reading I’m going to shorten “PCI DSS v1.2 Requirement 6.6” to just “PCI 6.6” since nearly everyone already refers to it as such.

Without further clarification, this requirement now compels an organization that stores, processes, or transmits cardholder data (“CHD”) to apply PCI 6.6 to all of their “web applications” that are publicly accessible, regardless of whether the applications participate in the storage, processing, or transmission of CHD. We must assume also that this requirement will be applicable regardless of other security controls as well, such as network segmentation. Disregarding all other changes to the PCI DSS, this one requirement could significantly increase the level of effort and cost of PCI assessments.

Before I get into advice for dealing with the new requirement, let’s examine the two options in relation to the change. First option is to have an application security organization perform security testing on the applications, correct identified vulnerabilities, re-test and repeat if necessary. Furthermore, such testing must occur “at least annually”, which should not be a huge problem…and “after any changes”, which could very easily force an organization to simply stop making any changes to its web applications. Unintended consequences.

The second option seems rather simple, straightforward, and innocuous unless you have actual experience with web application firewalls (“WAFs”). At a high level, WAFs have two basic modes of defense: 1. Blacklisting attack strings, typically pushed down by the vendor like anti-virus signature updates, and 2. Custom rule checking, which requires training the WAF to understand more of the application it is protecting. Both modes can often lead to false positives; the second and more effective mode requires a lot of human interaction throughout the system’s lifecycle. Add to the mix the large price tag for commercial WAFs and the new requirement that requires protection for all public facing web applications, and what you likely end up with is an organization implementing an easily defeated blacklist of your standard attack strings (XSS, SQL injection, etc). Unintended consequences

I do not want to be accused of PCI hate speech. In fact, I would like to sincerely applaud their efforts in helping to drive application security education and strongly encouraging organizations to address the problems. My concern, though, is that the more stringent the requirements become, the more likely organizations will continue (or begin) to do only the bare minimum in order to “check the box” instead of fixing the root cause of the problem. In Star Wars terms, “The more you tighten your grip, Tarkin, the more star systems will slip through your fingers.” –Princess Leia

With this information in mind, we now come to the issue of how to approach PCI 6.6. I’m going to suggest the most pragmatic approach that should work for most or all organizations, or at least those organizations that must endure PCI DSS assessments. In step-by-step format, naturally:

1. Classify applications – Separate all public facing web applications into a grid by user type (e.g. internal users only, external users only, mixed) and data sensitivity (e.g. CHD, public, internal/corporate, internal/sensitive).
2. Shorten list – If possible, move all web applications with internal users only to the internal network and provide external access via VPN. This should be an obvious move and something that should have been done well before PCI 6.6, but you might be surprised how often I see this situation.
3. Create assessment plan – First determine the budget for performing automated runtime vulnerability scans (the “cheap” option) against all remaining public facing web applications. Then the remaining budget—if any—can be reserved to add more effective assessment tasks to the most relevant applications, obviously starting with those that store, process, or transmit CHD.

As a general rule, I suggest using the results of an assessment against one application to identify and remediate potential weaknesses in other applications that either use the same codebase or were developed in similar fashion. This approach should lessen the number of findings in subsequent tests.

If all the advice in this post seemed familiar or common sense, then congratulations…you’ve been paying attention! The security industry has been evangelizing a risk-based approach for a long time now. Or at least some of us in the industry have been.