The Problems With Big Data

I developed a presentation on big data for a series of education sessions I am delivering for a financial institution trade association.  As I was putting the presentation together, I realized that this was probably a good topic for the blog as a lot of you are running headlong toward big data either for log data analysis or just as the next “I need to be doing this” technology fad.

Most people do not realize that big data has been around for quite a while, relatively speaking.  Google, Yahoo and similar Web service providers have been dealing with big data for years.  But it was only recently that big data management frameworks such as Apache Hadoop, Google BigQuery and MongoDB became publically available through shareware, commercial solutions and software as a service (SaaS).

So we are all on the same page, let us define “big data.”  The best definition I have seen for big data is from Gartner.

“Big data are high-volume, high-velocity, and/or high-variety information assets that require new forms of processing to enable enhanced decision making, insight discovery and process optimization.”

Examples of big data include information such as Web search results, electronic messages (e.g., SMS, email and instant messages), social media postings, pictures, videos and even system log data.  However, it can also include credit/debit card transactions, check images, receipts and other transactional information depending on the source of the information.  As a result, big data can easily end up in-scope for PCI compliance.

The first problem with big data is that organizations are expected to know the data going into their big data repositories.  The reason “know your data” is so important with big data is that it comes from potentially a wide variety of sources such as:

• Social media
• News feeds
• Images
• Streaming media (audio and video)
• Documents
• Messaging systems
• Audit logs
• Transaction logs
• Web sites
• System logs

With this diversity of information sources, it is anyone’s guess as to how much sensitive information could end up in an organization’s big data repositories.  But worse yet, anyone in the big data field will tell you that you need to anticipate all potential sensitive data so that you can secure and protect it appropriately.  Why is this?  Because big data tools allow everything to be searchable: text, images, audio, video, anything.  So if you do not want the information to be searchable, then you need to identify it and either encrypt it, truncate it or remove it so that it cannot be found.  As those of you that are using data loss prevention (DLP) or other tools to find cardholder data (CHD) stored on your systems are well aware, finding CHD is not as easy it would appear.  As a result, finding it in big data could be the ultimate finding the needle in the haystack game.

The next problem with big data is that the security tools for big data are very early in their development and, in some cases, are really bolt on after thoughts that use constantly running queries to find and protect the sensitive data.  While a lot of vendors claim they can secure data at the “field” level, I have spoken to a number of clients going through big data implementations that tell me this is a pipe dream at the moment.  As such, in all cases I am aware; big data protection is currently accomplished through totally encrypting all of the data and very severely restricting access.

Which begs the question, how can big data be PCI compliant?  Well it can be PCI compliant as long as: (1) access to it is extremely limited (only a very few people have access such as two to three), or (2) you are able to truncate, remove or encrypt the CHD contained in your big data hive(s).  Given that accurately locating CHD can be nearly impossible with current techniques, I just do not see option 2 as currently viable, so extremely limited access is your only workable option.  Even then I seriously doubt that big data is a good place for CHD to be stored as severely limiting access is also not viable given why big data is being implemented.  As a result, big data is probably not a good place for sensitive data, cardholder or otherwise, until the security tools catch up.

I think my readers will recognize why their log data would fit into the big data category.  It definitely has high volume as it typically comes from a large pool of devices that are generating potentially hundreds to thousands of entries per second which also satisfies the high velocity requirement as well.  And the high variety aspect is also satisfied as log data from a Cisco or Juniper device is nothing like log data generated from a Windows or Linux server or any other litany of network devices.  However, while the likelihood of CHD should be nonexistent in log data, CHD can still end up in log data due to debugging being performed.  As such, I am not certain I would be comfortable with log data in a big data hive until the maturity of the security tools is better.

The bottom line at the moment is I just do not see big data being ready for PCI compliance at this point.  I am sure that someday big data will be capable of being PCI compliant, but not at this time.  So all of you that need to be PCI compliant running towards the light at the end of the big data tunnel.  That light you see is not then end of the tunnel but an oncoming train about to run you over.

Pre-Authorization Data – The Card Brands Weigh In

I just wanted to pass along what I have heard thus far on this topic from the various card brands.  I’ll keep updating this post as I get more responses.

American Express (aka Trustwave) came back with the following.

“Per your inquiry regarding the storage of pre-authorized data, from what I can tell the American Express DSOP program currently does not address this topic. However, both the PCI SSC and the card brands (including American Express) have made it abundantly clear that pre-authorization data is to be protected with the same zeal as post-authorization data. That means encrypting it and restricting access to it. The reason the PCI SSC has not issued any directives regarding pre-authorization data yet is that it is a complicated environment and cannot be dealt with in a simple manner with the same approach working for all occurrences.

So, while pre-authorization data seemingly is not covered by the PCI DSS/DSOP programs at this time, you must do everything you can to protect it.

I hope this somewhat answers your question. I plan on doing more research to find any official documentation regarding the subject.”

MasterCard International followed up with their response.

“Merchants should talk with their acquirers to determine the pre-authorization rules particular to their vertical market and region.”

Visa quoted their Visa International Operating Regulations manual; section Account and Transaction Information Security Requirements VIOR 2.1.E (Updated):

“Ensure that all agents and Merchants do not store any of the following, subsequent to Authorization:
–    Full contents of any data taken from the Magnetic Stripe (on a Card, in a Chip, or elsewhere)
–    Card Verification Value 2 used to verify Card-Absent Transactions
–    PIN or the encrypted PIN block”

Discover has come back with the following.

“Currently the following restrictions apply:

(a) Sensitive Authentication Data obtained in connection with a Card Transaction must not be retained after receipt of an Authorization Response.
(b) Sensitive Authentication Data must not be recorded on Transaction Documentation or any other records or evidence of Card Transaction.
(c) Any processing, transmission, or storage of Cardholder Data must be conducted in accordance with PCI DSS requirements.”

JCB is still to be heard from.

I was deeply disappointed with MasterCard’s response.  Acquiring banks, for the most part, cannot answer basic questions about the PCI DSS, so we are supposed to believe that they are experts on retention of pre-authorization data based on a company’s vertical market and region?  Talk about passing the buck.

I am also surprised that I have yet to see any of the supposedly “very specific rules that prohibit any storage of sensitive authentication data (SAD) and do not make any exceptions.”  Not that I was expecting any as the Council has a tendency of making a mountain out of a molehill at times.

Based on the responses, I would highly recommend that merchants take American Express’ advice and protect pre-authorization data with the same voracity as they do with post-authorization data.

Is It ‘WHO’ Or ‘WHAT’ That Is Important?

There is a very active discussion going on in security circles about understanding adversaries and how that impacts security strategy.  I have taken a contrarian position in this argument and have stated that, in the scheme of things, I do not believe that you need to waste time understanding your enemy.  What I think matters most is what needs to be secured and how it needs to be secured.  This post is to discuss my rationale for this approach and relies on my prior post regarding the Fort Knox approach to security.

Sun Tzu famously said it was important to, “Keep your friends close and your enemies closer.”  The biggest difference with cyber-attacks is that the enemy are true mercenaries in that they come together because of an interest in a target, an interest in achieving their own particular goal, such as proving they are the best hacker or social engineer, or just because.  As a result, when your enemies can number in the hundreds or even thousands and have their own potentially unique motives for why they are attacking, it is near to impossible to do an analysis of the enemy, such as Sun Tzu suggests, that provides you with any sort of significant defensive advantage.

But what about advanced persistent threat (APT) attacks?  There is usually a common actor in APT, either a competitor, organized crime or a government.  However these sponsors usually hire the technical “muscle” for the actual attack.  The backer of the APT attack provides these mercenaries with a list of information they wish to be retrieved from the target organization(s).  So while APT can provide you with a traditional enemy, that enemy is obscured by the mercenaries actually conducting the attack.  Again, an analysis of the enemy provides limited to no advantage in your defense because you only see the mercenaries, not the sponsor.

But I think the biggest nail in the coffin for enemy analysis is related to attack strategies.  When reports from Verizon, Trustwave and other forensic examination firms consistently report that the same basic attack strategies are successful, it does not matter who the enemy is and why they are attacking when anyone from a neophyte to expert can break into your systems because of the same stupid mistakes or human errors.  By the time you have the enemy analysis done, your organization’s information is long gone.

In my opinion, ‘WHAT’ is more important in that organizations understand ‘WHAT’ information they need to protect and then go about appropriately protecting it.  If that sounds familiar, it should because that was the basis of my Fort Knox post.  If you think about it, a Fort Knox strategy does not worry about ‘WHO’ is trying to get the gold, it is all about protecting the gold regardless of ‘WHO’.

The bottom line is that in a cyber-attack, ‘WHO’ is attacking you is irrelevant.  You do not need to waste your time figuring out ‘WHO’ the attacker is and what are their motives.  It is all about your information that they wish to obtain.  So stop wasting time on enemy analysis and start properly protecting your organization’s critical, sensitive information.  I think you will find that the Fort Knox strategy will make your security efforts much more easy to implement and maintain.

UPDATE: In a brief moment of clarity on my part, I realized after making this post that the Fort Knox security approach is just another way of looking at the ‘Zero Trust’ security model that was proposed by John Kindervag of Forrester a while back.  See my earlier posts on the Zero Trust security approach for more information.

Zero Trust Security – The Cultural Discussion

Zero Trust Security – The Technical Discussion

2012 Database Threats

I attended a Webinar recently put on by Application Security Inc. regarding the threats to databases for the coming year.  If you did not attend it, you missed a good session.  But the most disturbing thing brought up was their top 10 list database vulnerabilities and misconfigurations.  Their top 10 list is:

1. Default or weak passwords
2. SQL injection
3. Excessive user and group privileges
4. Unnecessary DBMS features enabled
5. Broken configuration management
6. Buffer overflows
7. Privilege escalation
8. Denial of Service
9. Unpatched RDBMS
10. Unencrypted data

If you look at my post a while back on the 2011 Verizon Business Services’ reasons for why organizations were breached, there is a great correlation between Verizon’s report and what Application Security Inc. is saying.

Their first point about weak or default passwords is very clear and should not need to be discussed.  In this day and age, we should all be ashamed that this is even on the list, let alone the first item on the list.  The bottom line here is that, if you use default or weak passwords, you deserve to be breached.

They brought up and interesting point about SQL injection attacks that a lot of organizations do miss or underestimate and that is the internal SQL injection.  Most organizations are so focused on the external threat that they forget about the threat from the inside.  Worse yet, most security professionals and DBAs are unaware of the threat SQL injection poses even without the Web.  Since most of today’s attacks are perpetrated once past the perimeter, the protection from the inside attack is very relevant and very important.  Because once an attacker is on the inside, it is relatively trivial to use SQL injection or other techniques to obtain data.  More and more organizations are beginning to understand the insider threat and are firewalling all of their database servers away from the general user community as well as minimizing the number of users that have direct SQL access to those servers.

Excessive privileges cannot always be addressed at the DBMS level.  In today’s packaged software world, a lot of the rights are managed and maintained at the application level and that security matrix is maintained in a database table.  The granularity that can be granted is usually where things go awry because the application’s security system only provides an “all or nothing” approach.  Application vendors are getting better with this because of SOX, HIPAA, PCI and the like.  However, organizations typically need to be on the most current releases to have access to such enhanced security granularity.  Unfortunately, there are very few organizations that can afford the most current release or can implement the most current release due to their extensive modifications.  The simplest way to address this issue is the periodic reviews of database privileges and minimizing those users that have excessive privileges.  In the longer term, I expect we’ll see the return of the data dictionary with the addition of user rights and roles that will manage this problem.

Unnecessary features enabled are a vendor and DBA issue.  In some cases, vendors make changing features impossible or near to impossible once the RDBMS is installed.  In some cases, there are physical reasons as to why a feature must be enabled at installation.  However, there are also instances where features could be enabled or even disabled at anytime, but because the vendor only wanted to do that at installation, that is the only time you can deal with the feature.  This results in a lot of DBAs installing the RDBMS with every feature/function available, whether it is needed or not, just in case they might need it later on.  Do not get me wrong as I understand the drivers for this practice.  In today’s “I needed it yesterday” world, it is tough to be responsive when something will require an entire re-install of the RDBMS and migration of existing data in order to get something done.  It is time for IT people as a whole to start explaining to non-IT people that there are just some tasks that take time to do properly no matter how quickly anyone needs them completed.  Our infrastructure has become susceptible to attack in large part because of this rapid response desire.  If we intend to make things secure, we need to stop and think things through before creating even larger issues.

The pervious issue feeds directly into the next; broken configuration management.  Configuration management is broken because the vendors make it virtually impossible not to break it.  And even when configuration and changing configuration is easy and manageable, DBAs are not always as structured as other IT operational disciplines.  As a result, whether talking about the configuration of the RBDMS or the client that loads on the workstation, configurations are too broad because of the “just in case” factor.  I know it is a pain to only enable what needs to be enabled and then six months later have to reinstall everything just for a particular feature, but that is the correct way to do things if you intend to be secure.

Buffer overflows, privilege escalations and denial of service are all common vulnerabilities that organizations will have differing levels of success in mitigating.  I will tackle the easiest to address first, privilege escalation.  If there is any area where security can always be addressed it is with privilege escalation.  The reason privilege escalation exists is because someone, usually a developer, created the issue because they decided to allow users to perform a task that the user should not be allowed to perform.  Because if they were allowed to perform the function, then they would not need their privileges escalated to perform it.  The easiest thing to do is to disable those functions that require privilege escalation.  However, in some cases, that approach will create operational issues that will be unacceptable.  In those cases, monitor the daylights out of things so that you can be sure that the privilege escalation did not result in a different outcome.

In a lot of cases, there can be little done to address a denial of service (DoS) attack short of blocking the offender(s).  Denial of service does not compromise information; it just makes the information stored in the database unavailable.  And for most organizations, that is an important distinction.  If no information has been or can be compromised, then DoS is an annoyance and should be treated as such.  However, some DoS attacks can be used to defeat security measures in the RDBMS by causing the RDBMS to fallback to a basic operational state.  It is in these situations that one has to be careful because information can be lost.  The easy fix is to put a firewall in front of the database and enable DoS attack protections.

Buffer overflows are the most insidious attacks because, in some cases, there is little that can be done to stop them.  A lot of security professionals make the success of buffer overflow attacks sound like they are all the result of sloppy coding practices.  And while there is some truth to that view, the amount of which depends on the skills of your programmers, the success of buffer overflow attacks is also the result of embedding too much flexibility into our applications and leveraging the capabilities of the RDBMS.  In today’s world of open constructs such as SQL and RegEx, we have effectively made everyone a potential database programmer all in the sake of expediency.  Yes, customer service is highly improved, but at what cost?  Web application firewalls can minimize buffer overflows by “learning” how your SQL calls are typically structured, but they are not a complete answer nor do they completely remove the risks.  The way to fix the problem is to reduce functionality and make applications more complicated and difficult to use.  For most organizations that is not an option.  As a result, we must minimize the risks but be willing to accept the risks that remain as a result of our desire for ease of use and flexibility.  Minimizing the risk may mean implementing that Web application firewall internally as well as externally.

While I was glad to see that unpatched RDBMS software low on the top 10 list, I was very disappointed that it was still in the top 10.  One would think with all of the discussions about the importance of patching software, this would not occur in the top 10.  I understand the issues of compatibility and testing that make patching difficult, but really?  Maybe you need to invest in more than one or two instances of the RDBMS.  This is the cost of doing business the correct way.  If you are not doing things the correct way, then do not complain when you have a breach.  So while you saved yourself money on licensing costs on the front end, you likely paid for that cost savings a hundredfold on the back end.

I also understand the issues and fears with encryption.  For a lot of people, encryption is this mystical science that only certain “geeks” practice and practice well.  For others, the problem with encryption is the perceived loss of ready access to their data.  As time goes on, I would say that unencrypted data will rise to the top of the top 10 list.  Why?  Because the information age is all about the control of information.  The more information you control and can use to your advantage, the more power and control.  If your information can be readily obtained through public sources or the lax security surrounding your information systems, then you have little, if any, power or control.  The next 10 years will likely be spent by most organizations figuring out what information is critical to their business model and implementing the necessary protections around that information.  Critical information will be protected like the gold at Fort Know because, to that organization, that is their “gold” and it must be protected accordingly.  And that protection will likely involve encryption for some or all of it.

I know that people have a lot on their plates these days.  However, if you are a security person or a DBA, you need to leverage these surveys to your advantage and address the top 10 issues.  If more companies did this, the less data that would be breached.

Encryption Key Management Primer – Requirement 3.5

Before getting into key management, it is important to acknowledge that these requirements are not relevant to every encryption solution.  In the case of PGP, key management requires the user to create a private/public key pair and then authenticate using their passphrase.  It is not that the principles of key management do not apply to PGP; it is just that such key management practices are not usually associated with PGP.

Hardware security module (HSM) solutions typically provide facilities for strong key creation as well as a secure storage mechanism for keys.  A prime example of this concept is with the Trusted Platform Module (TPM) that is now standard on notebook computers.  I will try to point out the differences with these various approaches as I move through the discussion.

So with that as a background, let us start into our discussion of requirement 3.5.  The overall goal of requirement 3.5 is stated as:

“Protect any keys used to secure cardholder data against disclosure and misuse:  Note: This requirement also applies to key-encrypting keys used to protect data-encrypting keys—such key-encrypting keys must be at least as strong as the data-encrypting key.”

Typically, the first question I get is, “What in the world is a key-encrypting key?”  I need to push this question off, as there are other questions that need to be addressed before a meaningful discussion of the key-encrypting key (KEK) can occur.

With the key-encrypting key question put on the back burner, the second question most people ask is how does one accomplish protection of encryption keys?  In the case of PGP, that is the responsibility of the key owner not to disclose their passphrase.  It is the role of the HSM key repository to protect and manage keys.  If your encryption solution does not provide a secure repository for your encryption keys, then you need to provide such a capability.

When you have PGP or HSM, then protecting encryption keys is built-in.  A manual option is to store you encryption keys on paper or removable storage media and store the paper or removable media in a safe.  The problem with the manual option is that encryption keys are typically needed to boot the secure server or start an application that needs access to encrypted data.  The security surrounding the keys becomes problematic at best as operations personnel need regular access to the keys in order to keep systems and applications running.  As a result, most organizations desire to have their keys stored electronically for ease of access by servers and applications.

This now brings us to the concept of a KEK.  If you desire electronic storage of encryption keys, then you need to ensure their security.  If you store encryption keys in a folder on another server, you need to not only restrict access; you also need to encrypt them so that should someone copy the folder, they cannot read the keys.  To accomplish that, you encrypt the encryption keys wherever they are stored and then secure the key encrypting key (aka KEK).

This is where requirements 3.5.1 and 3.5.2 come into play.  These requirements are focused on ensuring that if an organization has taken the approach of managing their own encryption keys and electronically storing them, the keys are properly secured and managed by as few persons as feasible.

That does not mean that those of you using PGP or HSM are not responsible for also complying with requirements 3.5.1 and 3.5.2.  It is just that those of you using PGP or HSM likely have an easier time complying as both technologies can provide their own solutions for meeting these requirements.

In a future entry, I will discuss requirement 3.6.

Hashing Basics

I am catching some heat over the Encryption Basics post from some of my more ardent detractors that have called me on the carpet over making security and PCI “too simple” or “dumbed down.”  As I said in that post, it was in no way meant to be a full dissertation on the topic.  This post, as with the previous post, is not a complete dissertation either.  I just want to clarify some things so that people have a reasonable foundation on the topic of hashing.

Hashing comes up in requirement 3.4 which says:

“Render PAN unreadable anywhere it is stored (including on portable digital media, backup media, and in logs) by using any of the following approaches: one-way hashes based on strong cryptography (hash must be of the entire PAN), truncation (hashing cannot be used to replace the truncated segment of PAN), index tokens and pads (pads must be securely stored), strong cryptography with associated key-management processes and procedures.”

Before we get into hashing algorithms, I need to make sure everyone is clear that hashing is a one-way operation.  Unlike encryption, you cannot reverse data that has been hashed.  So if you are considering using hashing, you need to make sure that you will never need the original information such as a PAN.

Probably the most common one-way hash algorithm is SHA-1.  Unfortunately, it has been proved that SHA-1 has some issues that make it not as secure as its later iteration, SHA-2.  As such, those of you using SHA-1 should be migrating to SHA-2 or another hashing algorithm.  Those of you considering a solution that relies on SHA-1 should be asking the vendor if SHA-2 or another secure algorithm can be used.

The other most common one-way hashing algorithm is RSA’s Message Digest 5 (MD5) algorithm.  Like SHA-1, MD5 has also been deemed no longer acceptable due to a number of issues that make it no longer secure.  As a result, users of MD5 are recommended to migrate to RSA’s MD6, SHA-2 or another hashing algorithm.  And just so you are not confused, MD5 is still commonly used as a way to generate a checksum for confirming the validity of a download which is an entirely different purpose than what I am discussing here.

And to show you that things do not stand still in the cryptographic hashing world, the United States National Institute of Standards and Technology (NIST) is in the process of selecting the winner of their SHA-3 competition.  That winner will be announced sometime in late 2012.

While you can use a hash function “as is,” security professionals typical recommend the addition of a “salt” to complicate the result of the resulting hashed value.  A salt is two or more characters, usually non-displayable binary values, appended to the original value, in our example the PAN.  The salt adds a level of complexity to the resulting hashed value thus making the use of rainbow tables to break hashed values more difficult, if not impossible, to use.

One useful thing you can accomplish with a hashed value is that you can still use it for research as long as you are not using a rotating salt.  That means that the hashed PAN should have the same hashed value every time the same PAN is hashed.  This can be very important for fraud investigators and anyone else that needs the ability to research transactions conducted with the same PAN.  If these people do need the actual PAN, they will have to go to a different system that stores the encrypted PAN or go to their card processor for the PAN.

The PCI DSS issues a warning about using hashing and truncation together.  There is a note in requirement 3.4 that states:

“It is a relatively trivial effort for a malicious individual to reconstruct original PAN data if they have access to both the truncated and hashed version of a PAN.  Where hashed and truncated versions of the same PAN are present in an entity’s environment, additional controls should be in place to ensure that the hashed and truncated versions cannot be correlated to reconstruct the original PAN.”

This note is more relevant to situations where the truncation is the first six digits AND the last four digits.  However, it brings up a good point regarding all methods of information obscuring including encryption.  Never, ever store the obscured value along with the truncated value.  Always separate the two values and also implement security on the obscured value so that people cannot readily get the obscured value and the truncated value together without oversight and management approval.

Hopefully now we all have a base level knowledge of hashing.

Encryption Basics

NOTE: This is a revised version of my original post to reflect readers concerns regarding statements made that do not reflect best practices surrounding encryption key management.  A big thank you to Andrew Jamieson who reviewed and commented on this revised posting.

During the last couple of years, I have run into more and more questions regarding encryption and encryption key management than I thought existed.  As a result, I have come to the realization that, for most people, encryption is some mystical science.  The stories of the Enigma machine and Bletchley Park have only seemed to add to that mysticism.  Over the years, I have collected my thoughts based on all of the questions and developed this distilled and very simplified version of guidance for those of you struggling with encryption.

For the security and encryption purists out there, I do not represent this post in any way, shape, or form as the “be all, to end all” on encryption.  Volumes upon volumes of books and Web sites have been dedicated to encryption, which is probably why it gets the bad reputation it does as the vast majority of these discussions are about as esoteric as they can be.

In addition, this post is written in regards to the most common method of encryption used in encrypting data stored in a database or file and that is the use of an encryption algorithm against a column of data or an entire file.  It does not cover public key infrastructure (PKI) or other techniques that could be used.  So please do not flame me for missing your favorite algorithm, other forms of encryption or some other piece of encryption minutiae.

There are all sorts of nuances to encryption methods and I do not want to cloud the basic issues so that people can get beyond the mysticism.  This post is for educating people so that they have a modicum of knowledge to identify hyperbole from fact.

The first thing I want to clarify to people is that encryption and hashing are two entirely different methods.  While both methods obscure information, the key thing to remember is that encryption is reversible and hashing is not reversible.  Even security professionals get balled up interchanging hashing and encryption, so I wanted to make sure everyone understands the difference.

The most common questions I get typically revolve around how encryption works.  Non-mathematicians should not need to know how an encryption algorithm works, that is for the experts that develop and prove that they work.  In my opinion, unless you are a mathematician studying cryptography, I recommend that people trust the research conducted by the experts regarding encryption algorithms.

That is not to say you should not know strong cryptography from weak cryptography.  I am just suggesting that the underlying mathematics that defines a strong algorithm can be beyond even some mathematicians, so why we expect non-mathematicians to understand encryption at this level is beyond me.  My point is that the algorithms work.  How they work is not and should not be a prerequisite for management and even security professionals to using encryption.

This leads me to the most important thing people need to know about encryption.  If you only take away one thing from this post, it would be that strong encryption comes down to four basic principles.

• The algorithm used;
• The key used;
• How the key is managed; and
• How the key is protected.

If you understand these four basic principles you will be miles ahead of everyone else that is getting twisted up in the details and missing these key points.  If you look at PCI requirement 3, the tests are structured around these four basic principles.

On the algorithm side of the equation, the best algorithm currently in use is the Advanced Encryption Standard (AES).  AES was selected by the United States National Institute of Standards and Technology (NIST) in 2001 as the official encryption standard for the US government.  AES replaced the Data Encryption Standard (DES) that was no longer considered secure.  AES was selected through a competition where 15 algorithms were evaluated.  While the following algorithms were not selected as the winner of the NIST competition, Twofish, Serpent, RC6 and MARS were finalists and are also considered strong encryption algorithms.  Better yet, for all of you in the software development business, AES, Twofish, Serpent and MARS are open source.  Other algorithms are available, but these are the most tested and reliable of the lot.

One form of DES, Triple DES (3DES) 168-bit key strength, is still considered strong encryption.  However how long that will remain the case is up for debate  I have always recommended staying away from 3DES 168-bit unless you have no other choice, which can be the case with older devices and software.  If you are currently using 3DES, I would highly recommend you develop a plan to migrate away from using it.

This brings up another key take away from this discussion.  Regardless of the algorithm used, they are not perfect.  Over time, encryption algorithms are likely to be shown to have flaws or be breakable by the latest computing power available.  Some flaws may be annoyances that you can work around or you may have to accept some minimal risk of their continued use.  However, some flaws may be fatal and require the discontinued use of the algorithm as was the case with DES.  The lesson here is that you should always be prepared to change your encryption algorithm.  Not that you will likely be required to make such a change on a moment’s notice.  But as the experience with DES shows, what was considered strong in the past, is no longer strong or should not be relied upon.  Changes in computing power and research could make any algorithm obsolete thus requiring you to make a change.

Just because you use AES or another strong algorithm does not mean your encryption cannot be broken.  If there is any weak link in the use of encryption, it is the belief by many that the algorithm is the only thing that matters.  As a result, we end up with a strong algorithm using a weak key.  Weak keys, such as a key comprised of the same character, a series of consecutive characters, easily guessed phrase or a key of insufficient length, are the reasons most often cited as why encryption fails.  In order for encryption to be effective, encryption keys need to be strong as well.  Encryption keys should be a minimum of 32 characters in length.  However in the encryption game, the longer and more random the characters in a key the better, which is why you see organizations using 64 to 256 character long random key strings.  When I use the term ‘character’ that can be printable characters of upper and lower case alphabetic as well as numeric and special characters.  But ‘character’ can also include hexadecimal values as well if your key entry interface allows for hexadecimal values to be entered.  The important thing to remember is that you should ensure that the values you enter for your key are as hard to guess or brute force as maximum key size of the algorithm you are using.  For example, using a seven character password to generate a 256 bit AES key does not provide for the full strength of that algorithm.

This brings us to the topic of encryption key generation.  There are a number of Web sites that can generate pseudo-random character strings for use as encryption keys.  To be correct, any Web site claiming to generate a “random” string of characters is only pseudo-random.  This is because the character generator algorithm is a mathematical formula and by its very nature is not truly random.  My favorite Web site for this purpose is operated by Gibson Research Corporation (GRC).  It is my favorite because it runs over SSL and is set up so that it is not cached or processed by search engines to better guarantee security.  The GRC site generates 63 character long hexadecimal strings, alphanumeric strings and printable ASCII strings, not numerical strings provided by other random and pseudo-random number generator sites.  Using such a site, you can generate keys or seed values for key generators.  You can combine multiple results from these Web sites to generate longer key values.

In addition, you can have multiple people individually go to the Web site, obtain a pseudo-random character string and then have each of them enter their character string into the system.  This is also known as split key knowledge as individuals only know their input to the final value of the key.  Under such an approach, the key generator system asks each key custodian to enter their value (called a ‘component’) separately and the system allows no key custodian to come into contact with any other custodian’s component value.  The key is then generated by combining the entered values in such a way that none of the individual inputs provides any information about the final key.  It is important to note that simply concatenating the input values to form the key does not provide this function, and therefore does not ensure split knowledge of the key value.

Just because you have encrypted your data does not mean your job is over.  Depending on how your encryption solution is implemented, you may be required to protect your encryption keys as well as periodically change those keys.  Encryption key protection can be as simple as storing the key components on separate pieces of paper in separate, sealed envelopes or as high tech as storing them on separate encrypted USB thumb drives.  Each of these would then be stored in separate safes.

You can also store encryption keys on a server not involved in storing encrypted data.  This server should not be any ordinary server as it needs to be securely configured and very limited access.  Using this approach is where those key encryption keys (KEK) come into play.  The way this works is that each custodian generates a KEK and encrypts their component with the KEK.  Those encrypted components can then be placed in an encrypted folder or zip file where computer operations have the encryption key.  This is where you tend to see PGP used for encryption as multiple decryption keys can be used so that in an emergency, operations can decrypt the archive and then the key custodians or their backups can decrypt their key components.

Finally, key changes are where a lot of organizations run into issues.  This is because key changes can require that the information be decrypted using the old key and then encrypted with the new key.  That decrypt/encrypt process can take days, weeks even years depending on the volume of data involved.  And depending on the time involved and how the decrypt/encrypt process is implemented, cardholder data can potentially be decrypted or exposed because of a compromised key for a long period of time.

The bottom line is that organizations can find out that key changes are not really feasible or introduce more risk than they are willing to accept.  As a result, protection of the encryption keys takes on even more importance because key changes are not feasible.  This is another reason why sales of key management appliances are on the rise.

That is encryption in a nutshell, a sort of “CliffsNotes” for the non-geeky out there.  In future posts I intend to go into PKI and other nuances to encryption and how to address the various PCI requirements in requirements 3 and 4.  For now, I wanted to get a basic educational foundation out there for people to build on and to remove that glassy eyed look that can occur when the topic of encryption comes up.

Google Wallet

Good news for anyone considering using Google Wallet.  Google Wallet encrypts the PAN and only stores the last four digits of the PAN in clear-text according to a forensic assessment conducted by ViaForensics.  The other piece of good news from ViaForensic’s examination was that drive by attacks against Wi-Fi or near field communications (NFC) to intercept Google Wallet communications appear to fail.

Based on ViaForensic’s analysis, it appears that Google Wallet would likely comply with the PA-DSS.  The full PAN is encrypted and communication of the PAN via Wi-Fi or NFC is secured.  Granted there are a lot of other PA-DSS requirements that we do not have a window into that may or may not be PA-DSS compliant.  But on the whole, I would have to believe that Google Wallet would be PA-DSS certified.  So, why is Google Wallet not PA-DSS certified?

First and foremost, in the eyes of the PCI SCC and the card brands, Google Wallet and similar applications are storing pre-authorization data and are just an electronic representation of someone’s traditional wallet.  The PCI SSC does not certify traditional wallets, so why would they certify electronic wallets?  As a result, the PA-DSS does not apply.  Should Google and other vendors of electronic wallets ensure the security of cardholder data?  No doubt about it.

But a more important reason that the PCI SSC is backing away from certification is related to the other findings in ViaForensic’s report.  Their analysis also uncovered some not so good news in that Google Wallet stores a lot of personally identifiable information (PII) unencrypted.  This PII includes such information as cardholder name, expiration date, credit limit and account balance.  I think most people would now start to understand why the PCI SSC backed away from certifying such applications.

The PCI SSC does not want to be on the hook for the unsecured PII.  If the PCI SSC were to certify Google Wallet as PA-DSS compliant, I am sure their lawyers informed them that such a certification would drag them into lawsuits involving the exposure of the unsecured PII even though the PA-DSS does not cover PII outside of the PAN.  Their lawyers probably advised them that a PA-DSS certification would likely imply to users of these electronic wallet applications that their PII was included in the PA-DSS certification.  As a result, the PCI SSC and card brands would likely have to launch a massive and costly educational campaign to explain to the public that the PA-DSS certification was only related to protection of a cardholder’s PAN and nothing else.  And even with such a campaign, the PCI SSC would likely still get dragged into lawsuits over peoples’ PII being exposed.

And the likelihood of such lawsuits is very high.  Smartphones are regularly lost and the security protecting them is almost non-existent, if security is even enabled.  A hacker can easily take a lost smartphone and obtain enough information to perform identity theft.  Hackers could even decrypt the PAN given the high likelihood that the PIN to decrypt the PAN could be derived from other information on the smartphone.  The nightmare scenario would be development of malware delivered through the smartphone’s application store that harvests the PII.

At the end of the day, there is just too much risk involved in certifying such applications because there is just no way to manage the risk.  So those of you thinking the PCI SSC should certify these applications need to rethink your position.

FYI  This is my 200th post.  I never guessed that this blog would last this long and I want to take this time to thank all of you for keeping this going.

Merchant Beware – New Mobile Payment Solution Out In The Wild

Merchants need to be aware of a new mobile payment solution – Square from Square Inc.  A colleague pointed me to the Square site with the question, “Is this PCI compliant?”

Square appears to be a hardware/software solution for iPhones, iPads and Android devices.  It has a cute, square magnetic stripe reader for swiping cards, but also appears to provide the capability to manually enter cardholder data through these devices’ virtual keyboards.  This all appears to be similar to the iPhone that used to appear in the first Apple iPhone commercials that, for reasons that will become obvious, magically disappeared from their commercials very quickly and quietly.  It is also why Apple no longer uses iPhones or iPod Touches in their stores to process payments.

In referencing the PCI SSC’s PTS certification database, I could not find Square’s certification for the PTS standard.  Although, given the pictures on Square’s Web site, I really did not expect to find it certified to the PTS standard as there is no way it could meet the PTS standard.  Has Square submitted their solution for PTS certification?  It may have, but since the PCI SSC PTS certification database only lists those devices that have completed the certification process, there is no way for anyone to know if it has submitted Square until it is certified.  However, since the use of PTS certified devices is a requirement of all of the card brands, until Square is PTS certified, use of a Square device for processing of credit cards violates a merchant’s merchant agreement.  Game over.

While not complying with the PTS standard is a deal breaker in my opinion that is not the only PCI compliance issue.  In referencing the PCI SSC’s PA-DSS certification database, I could also not find the Square software application listed.  That situation was also not unexpected as the PCI SSC announced in a press release on June 24, 2011 that it was suspending the PA-DSS certification review of all mobile payment applications indefinitely.  As a result, there is no way Square’s software will be PA-DSS certified for the foreseeable future whether they submitted it for PA-DSS certification or not.  Not that the PA-DSS certification is a deal breaker for merchants to use the Square software, but it means that merchants using the Square software to process payments will have to have the Square software assessed to ensure it meets all of the PCI DSS requirements regarding payment applications.

And knowing what I know about all of these devices, I can guarantee that the Square software will not be PCI DSS compliant because all of these devices will store the cardholder data unencrypted for an untold amount of time until it is written over.  Even if Square’s software encrypts the data, the underlying OS will also collect the data in cleartext.  Forensic examinations of these devices have shown time and again that regardless of what the software vendor did, the data still existed in memory unencrypted.  And that unencrypted data in memory can exist in these devices for days, weeks to even months depending on transaction volume and other applications loaded on the device.  It is this surreptitious OS data collection activity, the security issues with other applications as well as other security concerns that caused the PCI SSC to suspend their PA-DSS certification activities of these applications.

There is only one solution that uses an iPhone or iPod Touch that is PTS and PA-DSS certified at this time and it is from Verifone.  The reason that Verifone’s PAYware solution is certified is that: (1) Verifone submitted it for the PCI certifications prior to the June 24 suspension and, the bigger reason in my book; (2) it relies on a digital back separate from the iPhone/iPod that performs the card swipe and all of the card data processing/transmission in a secure manner.  The iPhone or iPod Touch are used only as a display and cellular/Wi-Fi conduit for network connectivity.

The only other mobile payment solutions I am aware that are PTS compliant are purpose built credit card terminal using Wi-Fi or cellular communications.  These are considered terminals by the PCI SSC, so their underlying software is not required to be PA-DSS certified at this time, but they are required to be PTS certified.  In addition, these terminals have been in use in Europe for quite some time, so they are a proven secure solution.

The bottom line is that it is the merchant’s responsibility to ask vendors the right questions and weed out non-PCI compliant solutions.  The card brands and the PCI SSC are not in the business of regulating vendors, they leave that to the marketplace.

If you are looking for a PCI compliant mobile payment solution, talk to Verifone, Equinox, Ingenico or other recognized card terminal manufacturers as those are going to be your only PCI certified mobile payment processing options at this time.

UPDATE: I have had a number of people contact me about the certification status of the Square solution based on the fact that Square Inc. is listed on Visa USA’s Web site as a PCI DSS compliant service provider.  Remember, this listing only means that the card processing services provided by Square Inc. to their customers (i.e., all of the back office processing they do to process card transactions) are PCI DSS compliant, not that the devices that actually conduct the card swipe are certified PCI compliant.  A certified card terminal would be PCI PTS certified and the software running on the terminal would be PA-DSS certified.  The Square Inc. device that connects to iPhones and Android devices does not have that PCI PTS or PA-DSS certifications, therefore it should not be used for conducting credit card transactions.

UPDATE: Here is another solution that avoids the iPad altogether, Hubworks Interactive.  I spoke with them and their QSA and this product avoids the iPad by conducting the transaction in the digital framework surrounding the iPad.  The iPad is strictly used as a display and a Wi-Fi conduit.  The digital framework encrypts the cardholder data before it is ever seen by the iPad just as Mark Bower suggested would work.

UPDATE: July 25, 2012 – In the July issue of Transaction Trends, page 6, there is an announcement from Square that they are offering a new loyalty program for merchants using Square Register and Pay. Let us hope it is securely implemented as I am sure Square is using a customer’s PAN for tracking based on how they describe the program.

UPDATE: August 30, 2012 – The Square Web site is now implying that the transmission of card data is secured through “industry-standard encryption,” whatever that means.  There have been rumors that Square has implemented point-to-point encryption (P2PE) in their card readers but was not advertising that fact.  That is why I went to this page to see what, if anything, had changed.  However, based on the statements on this page, it appears that P2PE has not been implemented as you would think they would be touting that fact.

UPDATE: October 15, 2012 – A good friend of mine just started working at Square, so hopefully we can get to the bottom of how Square works and what risk there is to merchants using their solution.  Stay tuned.

More On Mobile Payments

As I have found out, the definition of “mobile payment” is defined by to whom you are talking.  For consumers, mobile payment means using their smartphone to pay for goods and services.  For merchants it includes the consumer definition as well as using smartphones or similar mobile devices to process payments.

Last year I wrote a post regarding mobile payments and the use of smartphones, primarily the iPhone, for use as credit card terminals.  When I wrote that first post, Apple was running an advertisement for the iPhone that showed it being used to process a credit card payment with the ubiquitous tag line, “There’s an app for that.”  Shortly after that post, the advertisement dropped the iPhone as a credit card terminal.  I am not aware that the PCI SSC or any of the card brands complained about that advertisement, but I found it interesting that those images of it processing a credit card were removed particularly given that a number of security and privacy issues that were and still are being discussed regarding the iPhone.

That is not to say that iPhone credit card adapters have not continued to be developed.  It is just that they are nothing like the one shown in that original Apple advertisement.  The first one that I came into contact with was Verifone’s PAYware Mobile solution and the fact that it is PA-DSS certified.  Whoa!  In my previous post I talked about all of the issues with the iPhone that make it almost impossible to be PCI certified.  How did Verifone create a PA-DSS certified application on the iPhone?  What Verifone did was to create a digital back to the iPhone.  All of the operations that need to comply with the various PCI standards are done through the digital back, not the iPhone.  The iPhone is just used as a display.  In the event that a credit card will not swipe through the digital back, the customer must go to a standard register.  I have also been privy to a number of similar iPhone applications.  All of them avoid the iOS interfaces as iOS is the problem in achieving PCI compliance.

While iPhone is the “Big Kahuna” of smartphones, it does not mean that Android and Windows Phone devices are not also used for credit card payments.  Unfortunately like the iPhone, Android and Windows Phone devices have similar issues that make them difficult, if not impossible; to have PA-DSS certified applications.  So from a merchant perspective, iPhone, Android and Windows Phone all have to be treated very carefully when they are used to process credit card payments.

But security concerns have not stopped merchants from rolling out mobile payments.  Starbucks recently introduced an iPhone and Android application that allows the customer to put their Starbucks cash card on their phone.  The application creates a 2D bar code with the cash card’s number.  The Starbucks POS system reads the bar code and automatically deducts the purchase from the account’s balance.  Within a week of releasing the application, it was determined that if you take a picture of the screen containing the bar code, anyone with the bar code can use the account until it cannot pay for a purchase.  So much for secure mobile payments.

If we expect to secure payments, the traditional credit card is just not going to get the job done.  EMV, aka Chip and PIN, is a short term technological fix but also a back up payment method for where I think we are really headed.  I truly believe that the future in payments is smartphones and other mobile devices with software that generate one-time transaction codes for paying for goods and services.  Whether those codes are displayed as a 15/16-digit number or bar code on a screen or transmitted via Wi-Fi, Bluetooth or RFID, a consumer will not need a traditional credit card.  A 15 or 16 digit number will be necessary to use so that POS systems do not have to be re-engineered to support the new payment method.  Scanners are already capable of reading bar codes from smartphone screens, so that much of the solution is already in place.  Wi-Fi, Bluetooth and RFID technology is coming as we speak so it is only a short matter of time before the infrastructure is in place to support such a solution.  All that is needed is the software.

Such an approach not only will secure card present transactions, but would also tackle the security issues we face with card not present transactions.  If done right, mobile payments can become the solution to our PCI compliance problem.