No matter how strong our encryption algorithm is, the safety of encrypted data is only as good as the safety of the secret key. We must choose keys that are hard to guess and we need to keep the key safe from attackers. On the other hand, we face disaster if we lose or forget the secret key that encrypts some critical data. The problem might be more than a personal one, especially if the encrypted information belongs to an employer or other proprietor. Encryption also has political implications because of its historic role in World War II and the Cold War.
Strong encryption is pointless unless the secret key is hard to guess. If the data is valuable and the attacker is highly motivated, use a relatively long and hard-to-guess passphrase for a workstation encryption key. For example, use a short piece of prose that an attacker is unlikely to guess. The point of a good passphrase is to substitute the difficulty of protecting the computer's data with the difficulty of protecting the passphrase. A good passphrase will be too difficult for attackers to guess even by using a dictionary attack.
A complementary problem, particularly with volume encryption, is that users will lose their data if they lose the secret key. Strong encryption is intended to resist attempts to retrieve the data through trial and error. If encrypted data can be retrieved by a legitimate user despite a forgotten password, then others may also retrieve this information regardless of the user's intentions.
A sensible strategy for file encryption is to use the same passphrase to encrypt all files that are likely to be shared by a particular group of people. For example, a user might write reports for several different projects. The user can encrypt all of the reports for one project with one passphrase and the reports for another project with a different one. The user shares the appropriate passphrase with the people associated with a particular project. This puts the minimum burden on everyone by minimizing the number of secrets to share. The author keeps secrets separate by using separate secrets.
Encryption also yields a situation in which people may need to keep a written copy of their memorized base secrets. In other access-control and authentication systems, users can usually ask an administrator to replace a lost or forgotten password. Recovery isn't so easy with encryption. If the user forgets the passphrase for strongly encrypted data, the data is effectively lost. For that reason, it is practical and even necessary to write down the passphrases for encrypted data. However, once the passphrase is written down, the paper containing it must be very carefully protected.
Encryption keys pose a dilemma because they do more than simply authenticate the authorized user: knowledge of the key also controls access to the encrypted data. This leads some proprietors and other authorities to want to treat encryption keys differently from passwords.
If the proprietor relies on passwords to authenticate people, then there is usually a rule against disclosing the password to anyone else. The proprietor can then hold the password's owner accountable for things done when the password gets used. Moreover, a strong rule against disclosing one's password also reduces the risk of social engineering attacks (Section 1.4). Users are less likely to fall for such trickery if they know with confidence that administrators will never ask for their password.
On the other hand, passwords used as encryption keys control the availability of the information, and arguably play a secondary role as far as accountability goes. Anyone who needs to retrieve the data must know the password that encrypts it. It doesn't really matter who uses the password as long as they're supposed to know it. This, in turn, leads some proprietors to demand that encryption keys always be shared with administrators to ensure availability of encrypted data.
Since memorized encryption keys are probably indistinguishable from passwords as far as most users are concerned, this produces an apparent contradiction in policy. One way proprietors can deal with this is training: make it clear what the difference is and why some passwords should be disclosed but not others. This is tricky, since most people won't understand the mechanisms to understand the differences. If users know they're supposed to disclose encryption passwords, then an attacker might manage to convince them that their logon passwords really are encryption passwords, too.
Key Escrow and Crypto Politics
Another way to provide separate access to encrypted data is to use key escrow or key recovery technology. Some encryption products provide a way by which the proprietor can supply an additional encryption key: this key will allow the proprietor to decrypt the data without having to know the key chosen by the data's creator. Thus, the creator can keep the key secret but the proprietor still has access to it.
Key escrow has a bad reputation in the computer security field because it evolved as part of the political controversy surrounding commercial encryption technology. Until DES was introduced, there was no strong encryption available to protect commercial or private telecommunications traffic. DES enabled the development of worldwide electronic financial transactions by protecting the transactions from sniffing and forgery.
However, DES also interfered with surveillance activities of the NSA and the FBI. In the United States, encryption technology was classified as a "munition" according to export control laws passed following World War II. Those laws were officially intended to restrict the development and sale of military encryption products, but they also applied to commercial products. Hardware and software systems using DES could be sold to overseas customers only after a special license was issued by the State Department.
As digital telecommunications flourished with the growth of the Internet, export restrictions on encryption technology became a major political issue. In the mid-1990s, the NSA developed the Escrowed Encryption Standard (EES), which became FIPS 185. EES incorporated secret encryption keys that would allow U.S. surveillance agencies to sniff encrypted messages. The best-known implementation was embedded in an integrated circuit called the "Clipper Chip." Few U.S. vendors or foreign customers showed much interest in the technology. The government tried to coerce vendors into implementing EES by exerting pressure through export regulations, but most vendors still refused. By the end of the 1990s, export restrictions on encryption products were mostly eliminated, and the EES was abandoned. But the distaste for escrowed encryption keys still lingers.
As a practical matter, key escrow can provide a valuable mechanism by which proprietors establish back-up keys for encrypted files or volumes. This reduces the risk of losing data by losing personally chosen encryption keys, especially if files are stored in encrypted form for a long period of time. Few people know how to ensure that an encryption key will remain secret but not get lost over the course of several years. Back-up keys, like those provided through key escrow mechanisms, may reduce the risk of losing the data by losing the key.
On the other hand, escrowed encryption doesn't make much sense in telecommunications systems. The data on the communications line is ephemeral; the valuable data is the information constructed at either end as a result of the communication. If a proprietor needs to eavesdrop on communications as part of regular business activities, there are much simpler ways to do it. Moreover, the key escrow mechanism proposed by the government involved sharing keys with third-party organizations: the surveillance agencies. We face greatly increased risks to our keys when we share them with larger and larger groups.