Townsend Security Data Privacy Blog

 A few months ago I wrote a definitive guide to Cryptographic Key Management. In it I wrote a section: A Brief History - the Need for Encryption Key Management. I wanted to expand upon the Classical Era of cryptography a bit because the story of data security goes back for millennia, and the twists and turns of this story can be felt even today.

Introduction

There has been a competition playing out through the centuries all the way from the highest corridors of power down to the shadiest back alleys. It is a struggle of those with a secret and those who want to uncover it. It is the story of cryptography and cryptanalysis.

As with every competition, each side is constantly trying to outfox the other. Peter Baofu described the competition this way, it is “the never ending cycle of replacing old broken designs” of cryptography and “new cryptanalytic techniques invented to crack the improved schemes.” In fact, “in order to create secure cryptography, you have to design against [all] possible cryptanalysis.” This means that both sides are in a never-ending arms race.

In his book, “The Future of Post-Human Mass Media,” Peter Baofu describes two main types of cryptanalysis: Classical and Modern Cryptanalysis. Let’s take a look at the Classical Period to see how this cat and mouse game has played out through the centuries:

The Classical Cat-and-Mouse Game

Classical Cryptography

One of the earliest forms of “secret writing” is the Substitution Cipher where each letter of the message is systematically replaced by another set of predetermined letters. In it’s most famous form, the Caesar Cipher, used by Julius Caesar himself (1st century, B.C.E):

“each letter in the plaintext is 'shifted' a certain number of places down the alphabet. For example, with a shift of 1, A would be replaced by B, B would become C, and so on.”

Another technique was Steganography, which literally means: “covered writing,” is the art of concealing a message in plain sight. Mehdi Khosrowpour recounts one of the first recorded instances (in the 5th century, B.C.E):

“Demaratus, a Greek who lived in Persia, smuggled a secret message to Sparta under the cover of wax.” It “ was to warn Sparta that Xerxes, the King of Persia, was planning an invasion ... by using his great naval fleet. He knew it would be very difficult to send the message to Sparta without it being intercepted. Hence, he came up with the idea of using a wax tablet to hide the secret message. In order to hide the secret message, he removed all the wax from the tablet, leaving only the wood underneath. He then wrote the secret message into the wood and recovered the tablet with the wax.”

Classical Cryptanalytic Response

While steganography is only hard to crack if you don’t uncover the message; substitution ciphers were meant to remain a secret even if the message fell into enemy hands. It remained a fairly reliable means of securing messages, so long as the cipher was not revealed.

All that changed with the first recorded technique of cryptanalysis: Frequency Analysis. This technique “can be traced back to the 9th-century [C.E.], when the Arabian polymath Abu Yusef Yaqub ibn Ishaq Al-Kindi (also known as ‘Alkindus’ in Europe), proposed in A Manuscript on Deciphering Cryptographic Messages.” It comes from the observation that certain letters appear more often than others in a given language (the letter “E,” for example, occurs most often in English). There also also common letter pairings (like “TH” in English).

So, in the case of the Caesar Cipher where the plaintext message is :

meet me at the theater

If each letter is shifted one letter in alphabet, it becomes:

nffu nf bu uif uifbufs

Frequency analysis would note that the most common letter in the ciphertext is “f” (which would suggest it is an “e”) and only letter pairing is “ui” (which would suggest the “u” is “t” and the “i” is “h”). If we replace these portions of the ciphertext we reveal:

_eet _e _t the the_te_

With these two facts of frequency analysis alone we have more than half the message deciphered. With a few logical leaps we could decipher the remaining the five letters. The simple substitution cipher was rendered useless.

The Classical Cryptography Counterattack

Over the centuries other ciphers were introduced like the Polyalphabetic Substitution Cipher where a repeating, offset key is used to encrypt the plaintext (see picture, courtesy of the Library of Congress). First perfected by Johannes Trithemius in 1518 (although other variants existed beforehand), the person encoding the message would switch alphabets for each letter of the message.

So, “meet me” would now become: “lcbp gy,” a ciphertext that simple frequency analysis could not break since most of the letter and pairing statistics of a given language are not easily recognized.

Although, in time, this type of cryptography was broken by the likes of Charles Babbage using modular arithmetic, the existence of his cryptanalytic techniques remained a military secret for some years.

Final Thoughts

Fascinatingly, it was the use of math to break a cipher that led to our current arms race in data security. The use of math and algorithms to break cryptography means you need longer keys to encrypt the data and prevent a brute force attack; which, in turn, means you need faster computers to break the encryption; which, in turn, means you need longer keys; etc.

Unlike today, however, it took centuries to break a cipher back then. Now, it is just decades. From the Hebern Electric Super Code Cipher Machine in the 1920s, to the Enigma Machine of the 1930s and 40s, to the Data Encryption Standard (DES) of the 1970s and 80s, each seemed invincible until enhanced cryptanalytic techniques or greater computing power toppled it. Our current cryptography is reliable and secure, but quantum computers loom on the near horizon and their non-binary logic could brute force attack our current public key cryptography and make them insecure.

And so the arms race continues. Fortunately, NIST has already forecasted this threat and called for replacements to our current standards, well before it is a crisis.  

Key Management Interoperability Protocol (KMIP) is quickly becoming the industry standard for ensuring your product or software can communicate seamlessly with cryptographic key managers.  In fact, a study by the Ponemon Institute in 2013 reported on the state of encryption trends and found that “more than half of those surveyed said that the KMIP standard was important in cloud encryption compared with 42% last year.”  This is surprising since KMIP v1.0 was first ratified three short years earlier on October 1st, 2010!

How Did it All Start?

The first meeting held to start discussing the new set of standards was on April, 24th 2009 in San Francisco in conjunction with the RSA convention that year.  In attendance were representatives from RSA, HP, IBM, Thales, Brocade, and NetApp. Their initial scope was to “develop specifications for the interoperability of key management services with key management clients. The specifications will address anticipated customer requirements for key lifecycle management”

But why was KMIP necessary to begin with?  The short answer: more and more organizations were deploying encryption in multiple environments.  But with encryption comes the need to properly manage the encryption keys. With encryption increasing across multiple enterprise applications it became harder to easily manage the keys from the different enterprise cryptographic applications.  Better standards were needed to create uniform interfaces for the centralized encryption key manager.

Companies soon saw the benefits of adopting KMIP.  Both large and small organizations need their key management to work every time and need it to scale as their organization grows.  And while other work was done to address this issue, like OASIS EKMI, IEEE P1619.3,  and IETF Keyprov KMIP was designed to have a broader scope than it’s predecessors and give more comprehensive standards for the industry.

How Was KMIP Initially Received?

In 2010, KMIP debuted at RSA.  HP, IBM, and others demonstrated that their client programs using the KMIP version 1.0 protocol could “communicate securely with key management servers. The clients and servers [demonstrated] essential use cases such as generating cryptographic keys, locating existing keys, and retrieving, registering, and deleting keys.”

In 2011 at the RSA Conference major players like IBM, RSA, and HP demonstrated KMIP 1.0 compatibility with their client programs.  And again in 2012 and in 2013 even more companies like Thales, NetApp, and Townsend Security demonstrated KMIP compliance.  With all these prominent players becoming KMIP compatible, it was a major signal to the industry that KMIP was rapidly becoming the industry standard for interoperable communications for key managers.

How is KMIP Thought of Now?

Fast forward to 2014.  The The Storage Networking Industry Association (SNIA) announced a testing program for KMIP conformance for its members.  In their words, “By introducing the KMIP Test Program for the industry, we’re helping to encourage not only the adoption of enterprise–class key management, but a means for vendors to test for conformance and provide an assurance of interoperability and a layer of trust to their customers.”

At  OASIS’ Interoperability Showcase at RSA 2016 16 companies, including Townsend Security, demonstrated KMIP compatibility.  And with the likes of VMware, Oracle, Quantum, and many others  demonstrating KMIP compatibility, KMIP has become a dominant standard in key management interoperability.

Final Thoughts

Encryption is your last, best defense for data at rest.  But encryption is only as good as your key management.  If the key is exposed to hackers, the data is lost as well.  This is why key management standards like KMIP have already attracted considerable interest, and will continue to do so.  The ability to have a variety of vendor applications, platforms, and databases all able to communicate with a centralized key manager enhances the data security posture of the enterprise.  And this is what organizations should strive to achieve.

OASIS built the standard to address a broader scope of issues than what older industry standards addressed. But KMIP still is actively being matured by OASIS (we are on version 1.3) and we should expect to see further enhancements and revisions to the standard as well as broader industry adoption.  This should give us confidence that KMIP as a well-accepted, road-tested standard will continue to grow in industry popularity in years to come.

In an unprecedented October surprise, Wikileaks dumped thousands of emails onto the internet from the Democratic National Committee (DNC), most of them concerning Hillary Clinton’s presidential campaign.  Later, in defending this move, Wikileaks founder Julian Assange, in an interview with FOX News, “said a 14-year-old could have hacked into the emails of Hillary Clinton's campaign chairman,” reported the Daily Mail.  Assange later revealed in the interview that John Podesta’s, Hillary’s campaign chairman, password was 'password.'  Politifact has gone on to challenge that assertion, saying that “Podesta was using a Gmail account, and Google doesn’t allow users to make their passwords ‘password.’”

Whatever John Podesta’s password was, it has sparked a good deal of renewed interest in good password management.  And far be it from me to downplay this crucial bit of data security.  We still have a long way to go.  In fact, SplashData just completed their survey of over 5 million people’s passwords and found that over 10% of people still use the most commonly guessable passwords like:

• password
• 123456
• qwerty
• passw0rd
• Password1
• zaq1zaq1

If you use any of these, stop it. Now.

But if that is all that we learn from the hack and subsequent data breach, we have missed the lesson.  As far back as June of 2016, it was widely reported, by the likes of Brian Krebs and Jeremy Kirk, that the DNC was vulnerable to attacks do to systemic weaknesses in cybersecurity.  In fact, in Jeremy Kirk’s article, it was noted that a press assistant emailed everyone a new password after a recent breach (a strong password at that: 'HHQTevgHQ@z&8b6').  The irony is, some of the email accounts had been compromised.  The hackers needed only to open the email and use the new password.

Strong passwords are not enough to rebuff the efforts of hackers to gain entry and to render the data useless in case of a breach.  We need proven security measures in order to keep the data safe.  

The data security measures below reflect specific things you can do to secure your data-at-rest in general. While there are more more specific measures you can take for email servers, it is important to remember that organizations have sensitive data everywhere, not just in emails.  That being said, since even seemingly benign emails at the DNC can blow up into political controversy, they probably need to follow these along with more email specific recommendations.  Follow along to find some of the best methods your organization should be using today to better secure your data security posture.

Multi Factor Authorization

As we have already mentioned, usernames and passwords, by themselves, are not enough to authenticate users.  Truly strong passwords are hard to manage and remember.  And once a system is compromised, login credentials can be scraped with keyloggers, malware, or other such attacks.

You need an external verification process.  You need multi factor authentication (MFA). MFA has traditionally relied on verifying you by two of three ways:

• Something that you know (i.e.: username, password, challenge questions/responses, one-time-use code, etc.)
• Something that you have (i.e.: token, RFID cards or key fobs,  mobile phones, etc.)
• Something that you are (biometrics)

Each of these methods have their advantages and drawbacks. For example:

• Challenge Questions:

• PRO: do not require any physical equipment on the user side
• CON: do rely on the user’s memory, which can be fuzzy when it comes to precisely writing the correct response
• CON: are vulnerable to deduction through inspection of social media accounts, etc.
• CON: are “something you know” and so fall into the same category as login credentials, thereby not taking advantage of any other kind of authentication
• Physical Equipment: (like RFID cards and tokens)
• PRO: do not rely on a person’s memory
• CON: can be stolen or lost
• CON: require active device management from an administrator

One method of authentication that is gaining ground because of its ease of use is authentication that relies on OAuth (an open standard for authorization).  It does not rely on physical fobs (which can be lost) or an SMS text (which can be intercepted).  It, instead, relies on cryptographic code that generates a time specific one-time-use codes based on the user’s secret key and the time. Since the code operates simultaneously (and separately) on the user’s device (typically a mobile phone) and on an internal server, with no need for an internet connection; it greatly reduces downtime because of internet issues and hackers intercepting the one-time-use code.

Encryption

Strong, Advanced Encryption Standard (AES) encryption as put forward by NIST should be used to encrypt all sensitive customer and company data.  In 2001 NIST formally adopted the AES encryption algorithm.  Since then, it has been proven countless times to render the data useless in the event of a breach.  In fact, it would take the fastest supercomputer 375 x 1050 years to brute force AES encryption by running through all permutations of an AES 256-bit encryption key.  In comparison, the Sun will reach its Red Giant stage in 54 x 108 years, engulfing Mercury, Venus, and possibly Earth.  In other words, the Earth will be incinerated by the then rapidly expanding Sun before a hacker could effectively crack AES encryption through brute force.

The good news, AES encryption comes standard in most database’s native encryption libraries.  Along with those free versions, there are a number of commercial products that rely on AES encryption available.  So finding a way to secure your data with AES encryption will be fairly easy.  That being said, it is important to understand the development time and performance hits each solution takes. Native encryption libraries are generally free but take a bit of development time.  Commercial solutions take less time to deploy but many times are file/folder level encryption products and have performance hits because they take a longer to encrypt/decrypt than column level encryption products.

Centralized Encryption Key Management

As we mentioned, AES encryption is extremely difficult to brute force attack.  It’s strength lies in its ability to encrypt the data with a very long key (typically 256-bit). But it’s strength is also its weakness.  If your encryption key becomes known to a bad actor, your encrypted data becomes compromised.  That is why any encryption strategy worth its salt will include proper, centralized encryption key management.  

When defending your encryption key with full lifecycle key management, consider these things:

• The encryption keys should be logically or physically separated from the encrypted data.  This way, if the encrypted data is compromised, they will not be able to decipher it.
• The encryption keys should only be generated with a cryptographically secure pseudo-random number generator (CSPRNG).
• Restrict administrator and user access to the keys to the least amount of personnel possible.
• Create clear separation of duties to prevent improper use of the keys by database administrators.
• Manage the full lifecycle of the keys from key creation, activation, expiration, archive, and deletion.

For a more comprehensive view of encryption key management, please view the Definitive Guide to Encryption Key Management.

Real Time Log Monitoring

Forrester, in 2013, promulgated the cybersecurity model of “Zero Trust.”  In it, they put forward the motto: “never trust, always verify.”  By this, they mean that all users should be authenticated, restricted to the least amount of data possible, and verified that they are doing the right thing through real-time monitoring.  Of which, they advocate for:  

• Real Time Event Collection in which you collect and log all events, in real time.
• Event Correlation in which you analyze all events and narrow in on the ones that do not conform to expected patterns.
• Resolution Management in which you investigate all suspect behavior and either classify them as either benign or a possible threat for further investigation.

There are many Security Information Event Management (SIEM) tools available that accomplish this.  For more information, refer to Gartner’s SIEM Magic Quadrant to find the tools that fit your needs.

Final Thoughts

Defending data-at-rest is a never ending struggle of building robust defenses and continuous improvement.  But, it's not a question of if, but when, a data breach will happen.  And if the DNC data breaches taught us anything is that breaches can be embarrassing and costly.  Since  hackers are only growing more sophisticated in their techniques, it is incumbent upon us to respond in ever increasing levels of agility and sophistication of our own.

The old models of the high, guarded perimeter with complex passwords to gain entry are just not enough.  We need a higher degree of authentication, sensitive data rendered useless, and constant real-time monitoring of all traffic.  You data depends on it.

“There are only two types of data that exist in your organization: data that someone wants to steal and everything else.”

- Forrester Research

In 2013, Forrester released an outline of their proprietary “Zero Trust Model” of information security to The National Institute of Standards and Technology (NIST).  Their model seeks to change “the way that organizations think about cybersecurity,” execute on higher levels of data security, and all the while “allowing for free interactions internally.”

But, when looking to better secure your organization’s data security posture, it is good to start with what has changed.  In the report, Forrester concluded that the old network security model was that of “an M&M, with a hard crunchy outside and a soft chewy center.”  It is the idea of the hardened perimeter around the traditional, trusted datacenter.  This old model is fraught with vulnerabilities as the traditional model is not equipped to handle new attack vectors with IoT, workforce mobility, and data centers moving to the cloud. It is increasingly becoming outmoded and weak.

In it’s place must come a data security model that takes into account the current network landscape and its vulnerabilities.  Enter, Zero Trust.  It builds upon the notion of network segmentation and offers key updates all under the banner: "never trust, always verify."

Below are the three main concepts to Zero Trust.  Follow along as we break down the trusted/untrusted network model and in its place rebuild a new trust model.

Assume All Traffic is a Threat

The first rule of “never trust, always verify” is that all traffic within the network should be considered a potential threat until you have verified “that the traffic is authorized … and secured.” Let’s look at these two components:

• Authorized Traffic: Each end user should present valid (and up-to-date) login credentials (i.e. username and password) as well as authenticate themselves with multi factor authentication for each session logging into the network.  Usernames and passwords are not enough.  Only multi-factor authentication can reduce the risk of a hacker obtaining and misusing stolen login credentials.
• Secured Traffic: All communication, coming from inside and outside of the network, should be be encrypted.  It should always be assumed that someone is listening in.  Using SSH or TLS and keeping abreast of their potential vulnerabilities is the only way to reduce the risk of exposure.

Give Minimal Privileges

The only way to minimize the risk of employees, contractors, or external bad actors misusing data is to limit the access each user/role is given to the least amount of privileges possible.  With this, it is a forgone conclusion that all sensitive data is already encrypted and minimal privileges are given as to who can decrypt it.  We implement a minimal privileges policy so that “by default we help eliminate the human temptation for people to access restricted resources” and the ability for hackers to access a user’s login credentials and thereby have access to the entire network.

Role-based access control (RBAC) model, first formalized by David Ferraiolo and Richard Kuhn in 1992 and then updated under a more unified approach by Ravi Sandhu, David Ferraiolo, and Richard Kuhn in 2000 is the standard today.  It’s ability to restrict system access only to authorized roles/users makes it the ideal candidate for implementing this leg of Zero Trust.  While Zero Trust does not explicitly endorse RBAC, it is best game in town, as of today.  For a deeper dive, visit NIST’s PDF of the model.

Verify People are Doing the Right Thing

Once we have authenticated each user and restricted them to the least amount of data possible to adequately do their job, the last thing to do is “verify that they are doing the right thing” through logging and inspection.

Here is a short (and certainly not exhaustive) list of techniques used to inspect all events happening in your network.  

• Real Time Event Collection: the first step is to collect and log all events, in real time.
• Event Correlation: Next you need to analyze all of the events and narrowing in on the events that need greater scrutiny.
• Anomaly Detection: In a related move, you will want to identify the events that do not conform to the expected pattern and investigate further.
• Resolution Management: All events that do not meet the expected pattern should be investigated and either classified as benign or deemed a possible threat and given for further investigation.

Note: There are many tools available that accomplish these.  Please refer to Gartner’s Security Information Event Management (SIEM) Magic Quadrant to find the tools that may interest you.

Final Thoughts

It's not a question of if, but when, a data breach will happen. Hackers grow more sophisticated in their attacks and threaten everything from intellectual property to financial information to your customers Personally Identifiable Information (PII).  The old model of the high, guarded perimeter with the trusted, internal network no longer functions as a secure model.  Zero Trust offers a more comprehensive approach to today’s data security needs.  As you look to deploy this model, begin to seek out tools that will help you.  Here is a short list of some of the tools to consider:

• Log Collection Tools: Some platforms, like the IBM i, have proprietary formats, that are difficult for SIEMs to read.  Make sure your SIEM can fully collect all needed logs.  If it cannot, find or make a tool that will properly capture and send the logs onto your SIEM.
• SIEM Tools:  As mentioned earlier in the article, there are many good SIEM tools out there to help you collect, analyse, and monitor all events on your network.
• Encryption (data-in-flight): Fortunately, there are many open source protocols for secure communications like SSH and TLS.
• Encryption (data-at-rest): Advanced Encryption Standard (AES) encryption is ubiquitous in most platform’s native encryption libraries.  There are also a number of products that offer column level to folder/file level encryption.
• Centralized Key Management: The encryption you deploy is only as good and the level of protection you give to the encryption keys.  Therefore, robust encryption key management is a must.
• User Access Management: Managing privileges, credentials, and multi factor authentication can be a daunting task.  The more more you can automate this, the better.

In many cases, adopting this approach will not be about bolting on a few products onto your existing data security framework but completely renovating it.  Don’t let expediency force you to defend your data with only half measures.  Take a deep dive into Zero Trust’s approach and see where you may be vulnerable.