You should all be aware of the Heartland Payment Systems Breach that happened on Inauguration day - had it been a different day it would be a front page story, perhaps it will be a front page story today? This post is to share my thoughts (that are speculation) based on my experience in payment systems and security rather than to re-hash the Heartland Press Release.
Let the pundits come: I can hear it now, “Was Heartland PCI Compliant ?” (They were/are btw, the QSA was Trustwave and they are listed as a compliant service provide dated April 30th, 2008) If they were PCI compliant how did a breach occur ? It is important to note (again, and again) that Compliance != Security. Compliance is a snapshot in time, PCI is not based on an organization’s own risk assessment of their environment. It is a prescriptive list of general IT Controls to be used as a baseline for “better” security then those organizations that are not compliant with the intent of reducing the risk of breaches. But you can be compliant and not secure, security is a process and constantly striving towards that end (it is like driving to infinity you never get there), is the goal, not compliance itself. I also expect to see merchants use this as an excuse – why are we spending all of this money for PCI complaint when attackers can successfully attack Processors anyway?
Processors and Service Providers are Fort Knox: Processors process for thousands of merchants and handle a large volume of transactions. Processors are in the business of processing card data, they require card numbers to communicate over the payment systems interchange networks as well as to provide settlement, clearing , and authorization files (among others) – The data of these files and message formats have account numbers, track data, cvv2 data (only the authorization messages include the last two) in the clear -but transmission is typically over a private leased line, use file level encryption, or transport level encryption, but there is a place in the “PCI-ZONE” of companies that sends this data in clear across the network - making sniffing the traffic a threat here. Can this be further controlled by further network segmentation and control at processors? I think so (topic of another post). Only the Payment Switch would need to have direct connections to these end-points not the full PCI-Zone.
Blame the QSA: Nobody has blamed the QSA yet, and I don’t know if anyone will, but I’m sure someone will try to; when breaches occur they are going to be asked what they missed ? What does that mean for QSA’s or IT auditors ? How do your work papers look ? Can a independent third party look at your work papers as evidence and come to the same conclusion as you? Did you test to the control or did you test something else ? Did you understand intent of the control ?
Issuing Banks: You don’t get to hear much of the story about the burden on Issuing Banks here. They have the cost of notifying cardholders, postage expense, customer service calls, brand perceptions problems due to confused customers, costs to reissue cards, etc. This includes both credit, debit, payroll, gift and others.
Press Release: Two things that got me about the Press Release, one was its timing, which may or may not be a coincidence, and the second was the focus was on the data that was not compromised, not the data that was. Like why not add: The cardholders’ favorite colors and cereals were also not compromised Account Numbers, Expiration Date and Track Data (Track 1 contains Name, Track 2 doesn’t) were compromised, cloned cards can be made from these “dumps” of magstripe data.
PCI Compliance: Remember that Service Providers and Processors where the very first to be scrutinized under Visa’s CISP and MasterCard’s SDP programs, and later PCI. The truth is that many processors and gateways should have close to 5 years or so of experience with PCI compliance and reviews if they are not a new service provider. Also: If you ever read through some of the actual audit procedures of PCI - notice what the auditors actually test: some tests are just based on Inquiry or documentation alone, furthermore some tests that do not test the operating effectiveness of some of these controls to a period of time.
Cost of the Breach: Anything that you hear is a guess - nobody knows for sure– (Well the card brands and affected issuers probably will) - we know that Heartland does 100 million in volume a month, and can safely assume that data was sniffed for a few months (some reports it as in place as early as May 2008) – we don’t know the unique number of card numbers nor do we know which card numbers are affected. I would guess around $600 million - based on cost of record at ~$6.00 and assuming 100 million unique card numbers were affected.
How did it Happen: Let’s look at the PR:
“ After being alerted by Visa® and MasterCard® of suspicious activity surrounding processed card transactions, Heartland enlisted the help of several forensic auditors to conduct a thorough investigation into the matter. Last week, the investigation uncovered malicious software that compromised data that crossed Heartland’s network.”
There really isn’t enough information here but let’s make an educated guess: We know that malware/software was installed. So we have the threat of physical security/social engineering that would install a piece of software, an OS or Application level attack, or a targeted piece of malware that was installed in the PCI Zone accidentally. Perhaps a combination of some of these. This type of Malware (sniffers) requires administrator system or root level access in order to sniff network traffic in promiscuous mode (in most cases.) There are ways to flood network switches to make them act as a hub to broadcast traffic, and there are also VLAN hopping attacks.
PCI 1.2.1 / 1.3.5
Let’s look at PCI 1.2.1:
“Restrict inbound and outbound traffic to that which is necessary for the cardholder data environment”
and PCI 1.3.5
“Restrict outbound traffic from the cardholder data environment to the Internet such that outbound traffic can only access IP addresses within the DMZ”
So the data would need to be collected and re-transmitted to a drop site, the drop site was probably actually multiple drop sites, and I’m guessing a well known outgoing port such as 443/SSL was used with encrypted payloads. These two PCI controls are intended to prevent outbound access from the cardholder environment, did these systems have outbound Internet access ? or if they did was it controlled ? and if it was controlled was it able to pass though the allowed outgoing traffic ?
Anti-Virus / Firewalls / Encryption: These are terms that are used to define ones security, Security is not a product - But let’s look at each of these briefly:
Anti-Virus - The effectiveness of Anti-Virus is poor, if you are familiar with services such as VirusTotal or have read something like this <-- This is why not all malware can be detected by Anti-Virus.
Firewalls - I swear I think people think that firewalls are magic devices – Hollywood and TV Land don’t really help here either. Understand what a firewall does - it works at the network layer and can either allow or block IP Address or ranges and port numbers, that is basically all they do. Granted, Application level firewalls and Web Application Firewalls can inspect the content of the traffic (I’m not talking about these)
Encryption: There are different types of encryption, each protects different things in different ways.So you you say something to the effect of:
“We have industry leading encryption”
Are you talking about File Level Encryption ? Disk Encryption (which only works at data at rest), Transport encryption, encrypted data elements or application level encryption, column level or transparent encryption in a database ? Understand that Encryption is not Encryption is not Encryption. For example using a product to encrypt a disk to store data at rest, does not provide encrypted data elements and transport level encryption. And lastly End-to-End Encryption solutions would not of prevented this either: see my post here: When End-to-End Encryption is really not End-to-End.