Lessons for the Internet Age
Let's pause for a moment to consider some of the lessons of cryptographic history—morals that were well-understood by the early twentieth century. In the late twentieth century, cryptography changed drastically because of modern computer technology and new cryptographic algorithms, but these lessons are still true today. They are too often forgotten.
Breakthroughs Happen, but News Travels Slowly
Mary Stuart was beheaded when her letters plotting against Elizabeth were deciphered by frequency analysis, which Al-Kindi had described nine centuries earlier. Older methods have also remained in use to the present day, even for high-stakes communications. Suetonius explained the Caesar cipher in the first century A.D. Yet two millennia later, the Sicilian Mafia was still using the code. Bernardo Provenzano was a notorious Mafia boss who managed to stay on the run from Italian police for 43 years. But in 2002, some pizzini—ciphertexts typed on small pieces of paper—were found in the possession of one of his associates. The messages included correspondence between Bernardo and his son Angelo, written in a Caesar cipher—with a shift of three, exactly as Suetonius had described it. Bernardo switched to a more secure code, but the dominos started to topple. He was finally traced to a farmhouse and arrested in April 2006.
Even scientists are not immune from such follies. Although Babbage and Kasiski had broken the Vigenère cipher in the mid-nineteenth century, Scientific American 50 years later described the Vigenère method as "impossible of translation."
Encoded messages tend to look indecipherable. The incautious, whether naïve or sophisticated, are lulled into a false sense of security when they look at apparently unintelligible jumbles of numbers and letters. Cryptography is a science, and the experts know a lot about code-breaking.
Confidence Is Good, but Certainty Would Be Better
There are no guarantees that even the best contemporary ciphers won't be broken, or haven't been broken already. Some of the ciphers have the potential to be validated by mathematical proofs, but actually providing those proofs will require deep mathematical breakthroughs. If anyone knows how to break modern codes, it is probably someone in the National Security Agency or a comparable agency of a foreign government, and those folks don't tend to say much publicly.
In the absence of a formal proof of security, all one can do is to rely on what has been dubbed the Fundamental Tenet of Cryptography: If lots of smart people have failed to solve a problem, then it probably won't be solved (soon).
Of course, that is not a very useful principle in practice—by definition, breakthroughs are unlikely to happen "soon." But they do happen, and when they do, indigestion among cryptographers is widespread. In August 2004, at an annual cryptography conference, researchers announced that they had been able to break a popular algorithm (MD5) for computing cryptographic operations called message digests, which are fundamental security elements in almost all web servers, password programs, and office products. Cryptographers recommended switching to a stronger algorithm (SHA-1) but within a year, weaknesses were uncovered in this method as well.
A provably secure encryption algorithm is one of the holy grails of computer science. Every weakness exposed in proposed algorithms yields new ideas about how to make them stronger. We aren't there yet, but progress is being made.
Having a Good System Doesn't Mean People Will Use It
Before we explain that unbreakable encryption may finally be possible, we need to caution that even mathematical certainty would not suffice to create perfect security, if people don't change their behavior.
Vigenère published his encryption method in 1586. But foreign-office cipher secretaries commonly avoided the Vigenère cipher because it was cumbersome to use. They stayed with simple substitution ciphers—even though it was well-known that these ciphers were readily broken—and they hoped for the best. By the eighteenth century, most European governments had skilled "Black Chambers" through which all mail to and from foreign embassies was routed for decryption. Finally, the embassies switched to Vigenère ciphers, which themselves continued to be used after information about how to crack them had become widely known.
And so it is today. Technological inventions, no matter how solid in theory, will not be used for everyday purposes if they are inconvenient or expensive. The risks of weak systems are often rationalized in attempts to avoid the trouble of switching to more secure alternatives.
In 1999, an encryption standard known as WEP (Wired Equivalent Privacy) was introduced for home and office wireless connections. In 2001, however, WEP was found to have serious flaws that made it easy to eavesdrop on wireless networks, a fact that became widely known in the security community. Despite this, wireless equipment companies continued to sell WEP products, while industry pundits comforted people that "WEP is better than nothing." A new standard (WPA—Wi-Fi Protected Access) was finally introduced in 2002, but it wasn't until September 2003 that products were required to use the new standard in order to be certified. Hackers were able to steal more than 45 million credit and debit card records from TJX, the parent company of several major retail store chains, because the company was still using WEP encryption as late as 2005. That was long after WEP's insecurities were known and WPA was available as a replacement. The cost of that security breach has reached the hundreds of millions of dollars.
Similarly, many of today's "smart card" systems that use RFID (Radio Frequency Identification) tags are insecure. In January 2005, computer scientists from Johns Hopkins University and RSA Data Security announced that they had cracked an RFID-based automobile anti-theft and electronic payment system built into millions of automobile key tags. They demonstrated this by making multiple gasoline purchases at an Exxon/Mobile station. A spokesman for Texas Instruments, which developed the system, countered that the methods the team used were "wildly beyond the reach of most researchers," saying "I don't see any reason to change this approach."
When encryption was a military monopoly, it was possible in principle for a commander to order everyone to start using a new code if he suspected that the enemy had cracked the old one. The risks of insecure encryption today arise from three forces acting in consort: the high speed at which news of insecurities travels among experts, the slow speed at which the inexpert recognize their vulnerabilities, and the massive scale at which cryptographic software is deployed. When a university researcher discovers a tiny hole in an algorithm, computers everywhere become vulnerable, and there is no central authority to give the command for software upgrades everywhere.
The Enemy Knows Your System
The last lesson from history may seem counterintuitive. It is that a cryptographic method, especially one designed for widespread use, should be regarded as more reliable if it is widely known and seems not to have been broken, rather than if the method itself has been kept secret.
The Flemish linguist Auguste Kerckhoffs articulated this principle in an 1883 essay on military cryptography. As he explained it,
- The system must not require secrecy, and it could fall into the hands of the enemy without causing trouble.... Here I mean by system, not the key itself, but the material part of the system: tables, dictionaries, or whatever mechanical apparatus is needed to apply it. Indeed, it's not necessary to create imaginary phantoms or to suspect the integrity of employees or subordinates, in order to understand that, if a system requiring secrecy were to find itself in the hands of too many individuals, it could be compromised upon each engagement in which any of them take part.
In other words, if a cryptographic method is put in widespread use, it is unrealistic to expect that the method can remain secret for long. Thus, it should be designed so that it will remain secure, even if everything but a small amount of information (the key) becomes exposed.
Claude Shannon restated Kerckhoffs's Principle in his paper on systems for secret communication: "... we shall assume that the enemy knows the system being used." He went on to write:
- The assumption is actually the one ordinarily used in cryptographic studies. It is pessimistic and hence safe, but in the long run realistic, since one must expect his system to be found out eventually.
Kerckhoffs's Principle is frequently violated in modern Internet security practice. Internet start-up companies routinely make bold announcements about new breakthrough proprietary encryption methods, which they refuse to subject to public scrutiny, explaining that the method must be kept secret in order to protect its security. Cryptographers generally regard such "security through obscurity" claims with extreme skepticism.
Even well-established organizations run afoul of Kerckhoffs's Principle. The Content Scrambling System (CSS) used on DVDs (Digital Versatile Disks) was developed by a consortium of motion picture studios and consumer electronics companies in 1996. It encrypts DVD contents in order to limit unauthorized copying. The method was kept secret to prevent the manufacture of unlicensed DVD players. The encryption algorithm, which consequently was never widely analyzed by experts, turned out to be weak and was cracked within three years after it was announced. Today, CSS decryption programs, together with numerous unauthorized "ripped" DVD contents, circulate widely on the Internet (see Chapter 6, "Balance Toppled" for a more detailed discussion of copy protection).
Kerckhoffs's Principle has been institutionalized in the form of encryption standards. The Data Encryption Standard (DES) was adopted as a national standard in the 1970s and is widely used in the worlds of business and finance. It has pretty much survived all attempts at cracking, although the inexorable progress of Moore's Law has made exhaustive searching through all possible keys more feasible in recent years. A newer standard, Advanced Encryption Standard (AES), was adopted in 2002 after a thorough and public review. It is precisely because these encryption methods are so widely known that confidence in them can be high. They have been subjected to both professional analysis and amateur experimentation, and no serious deficiencies have been discovered.
These lessons are as true today as they ever were. And yet, something else, something fundamental about cryptography, is different today. In the late twentieth century, cryptographic methods stopped being state secrets and became consumer goods.