Cryptography Operational Protocols and Algorithms

Leaving aside for a moment the issues of protocols and algorithms, the first choices to be made are the cryptographic strength of the system, embodied by choices of algorithm and key length, and some of the key-management questions. It is most important to choose truly random keys that are sufficiently long, keep those keys secret, and change keys "often enough." Again, this all supposes that you have selected a good cryptosystem; all the security of the system lies in the keys, and none in the algorithm itself.

We will consider key randomness and key secrecy shortly. For now, let us consider the selection of key length and the frequency of key updates.

Key Length

Given a reasonably strong algorithm, how well the data is protected depends largely on the length of the encryption key. Fundamentally, an encrypted message must remain secret for the useful life of the information. To a large extent, the value of the information in the encrypted message will govern the resources used to attack it. For example, an attacker would be foolish to spend $1 million to obtain information worth $1,000, but he might spend $1 million to obtain a secret worth $2 million. Here are some examples.

• Financial credentials must remain secret beyond their validity period.
• Contract bids must remain secret beyond the contract award.
• Editorial material must remain secret until published.
• Confidential personal information must remain secret beyond the lifetime of the individual.

Today, it is common to use 128-bit keys for symmetric algorithms, both for communications security and for the security of data to be protected for 20 years. The necessary key lengths for public-key algorithms vary considerably. The current recommendation for the RSA public-key algorithm, for example, is to use a minimum length of 1024 bits, with 2048 bits used for especially sensitive applications or longterm keys.

Key Updates

Cryptographic keys do not last forever; they need to be updated from time to time. The proper lifetime of a key is a function of the value of the items encrypted, the number of items encrypted, and the lifetime of the items encrypted. We have already discussed lifetime. If a key can be broken by a properly equipped adversary in 2 years, and the lifetime of information encrypted using the key is 6 months, then the key should be changed at least every 18 months so that an attack mounted on the first item encrypted will not succeed until after the last item encrypted loses its value.

The number of items encrypted is an issue for two reasons. First, if individual encrypted items have a market value, then the sum of the values of all encrypted items is the proper measure against which to balance the resources an attacker may bring to bear. Second, some cryptosystems can be attacked more easily when a large body of ciphertext is available. This effect is more difficult to quantify, but again, it is a good idea not to use a key for too long.

Another factor that leads to frequent key updates is paranoia. The longer a key has been in use, the greater the chance that someone has compromised the key storage system and obtained the key by subterfuge rather than by brute force attack.

It is important to note that changing a key does not increase the time that an attacker will need to find it using brute force or any other method of cryptographic attack. Changing keys does, however, limit the amount of information revealed if any particular key is found. For example, if the encryption keys are changed every month, then only one month's worth of information is disclosed if a key is discovered.

Perfect Cryptosystem One-Time Pads

Is there a perfect cryptosystem? Surprisingly, the answer is yes. It is called the onetime pad. The idea of the one-time pad is to have a completely random key that is the same length as the message. The key is never reused, and only the sender and the receiver have copies. To send, for example, a 100-bit message, the message is exclusiveORed' with 100 bits of the key. That portion of the key is crossed off, never to be used again. The receiver reverses the process, exclusive-ORing the ciphertext with her copy of the key to reveal the message. If the one-time pad key contains truly random bits, this scheme is absolutely secure. The attacker does not know what is on the pad and must guess—but there is no way to know when he is right. By changing the guess, the attacker can decode the ciphertext into any message, be it "attack at dawn" or "negotiate surrender."

The one-time pad offers perfect security and is indeed used when perfect security is needed, but the system has many disadvantages.

• The pad must be truly random. Any structure at all can be used to break the system. Creating truly random characters is difficult, and creating a vast quantity of them is more difficult.
• The pad must never be reused. If a sheet is used twice, then the two sections of ciphertext encrypted using the same page can be compared, possibly revealing both.2 Since the pad is consumed as messages are sent, the pad has to be very long or frequently replaced.
• The pads must be distributed, stored, and used with absolute secrecy. Because the ciphertext cannot be successfully attacked, the obvious point of attack is to copy or substitute a pad.
• Every pair of correspondents must have a unique pad, leading to immense practical difficulties of distribution.

These practical difficulties effectively restrict the use of one-time pad systems to situations in which cost is no object. For most other situations, cryptosystems are used in which the length of the key is fixed, and the key can be attacked by exhaustive search.

Computers and Privacy part 3

RSA Public Key Cryptography

Several algorithms are used today for public-key asymmetric encryption. The most widely known is called the RSA algorithm, named after its inventors, Ronald Rivest, Adi Shamir, and Len Adleman. This method is based on multiplying two prime numbers to come up with the key pair. Further mathematical functions are performed after the multiplication to create the actual key pair. It's a simple matter to use a computer to come up with a rather large prime number, but it's a difficult computation task to take the result of this multiplication and the subsequent operations performed and determine which two prime numbers were used to generate it.

If you want to learn more about RSA, visit the Web site for the company founded to market this technology: www.rsasecurity . com. The RSA Security Web site is an excellent resource for encryption techniques overall, but also has a lot of information pertaining to the RSA algorithm, which has been licensed to a large number of software and hardware security providers.

Because of the difficulty in cracking RSA-encrypted data, it has been adopted by a large number of vendors, including Sun, Microsoft, and Novell, and is the most widely used cryptosystem today.

Digital Certificates

Digital certificates are used to bind a person's name (or an identity) to a public key. Certificates, then, must come from a trusted authority. The certificate itself is determined to be valid (that is, it was issued by the certificate authority [CA] it claims to represent) by a digital signature. Because the public key of a CA can be known to anyone, it is a simple computational matter to use the CA's public key to determine that the digital signature is valid. After this is done, the certificate itself can be assumed to contain a valid identity (a user, a corporation, or another entity) associated with a public key. Using a digital certificate, you then can obtain the public key for a person and use it to encrypt data to be sent to that person, who then can use his own private key to read your message.

CAs can be trusted companies on the Internet, or you can act as your own CA in your company. Included with Windows 2000 Advanced Server and the family of Windows 2003 servers, for example, is Microsoft's Certificate Services, which can be used within a company that wants to manage its own digital certificates. If you have branch offices and want to use digital certificates to certify public keys used for communicating over the Internet, you can set up your own certificate servers in your enterprise. Or you can use a commercial company (such as VeriSign) and obtain certificates from a third party.

In practice, it also is possible for a hierarchy of certificate servers to be set up, with a single root server being the most trusted certificate server in your enterprise. Then, child certificate servers are created, which can be validated by the end user because the child certificate server itself has a Certificate from the root server (or another server in the hierarchy leading back to the root server) that validates its certificate. It's all a game of trust, however. If the secret key of the root server's key pair becomes compromised, it's possible to impersonate the certificate server and all security is lost. Most certificates also are issued with an expiration date, which can be used to ensure that new certificates, created using a new key pair, are in use.

For this reason, should you choose to operate your own certificate server(s) in your network, you need to take extreme security precautions to safeguard the private key. Likewise, if you use a third- party commercial certificate service, you need to read the policy of that company to determine how it verifies the identity of the end users that it issues certificates to. For example, a CA might simply verify the email address of the requestor and issue a certificate. For a software publisher, the CA might conduct some kind of background check and require further evidence before it issues certificates to the company. Before you decide to use a commercial service for issuing digital certificates, be sure you investigate the company's policies for both issuing and revoking certificates.

Pretty Good Privacy (PGP)

One of the most popular encryption programs on the Internet for a number of years now has been PGP, originally developed by Phillip Zimmerman. PGP uses public-key cryptography and has been ported to many computer platforms, including Unix, Linux, and, of course, all versions of Windows from Windows NT and Windows 95 onward.

PGP Corporation currently markets the commercial version of PGP. PGP is available in a Universal Series version for network gateways, Whole Disk Encryption for desktop and enterprise systems, command-line for servers and mainframes, and home and professional desktop versions. PGP also maintains a Global Directory of PGP keys, replacing the PGP Keyserver service. A 30-day trial of the home desktop version is available.

PGP has been established as an Internet proposed standard through the Request for Comments (RFC) process. RFC 2440, "OpenPGP Message Format," was written in 1998 and details the specification.

An international site devoted to PGP also can be used to download PGP. Visit the PGPi Project International PGP home page at www pgpi. org/ to learn more about PGP International. The downloads available from this site include support for the following platforms:

Amiga Atari BeOS EPOC (Psion, and so on) MacOS

Newton OS/2 PalmOS Unix MS-DOS

Windows 3.x Windows 95/98/NT Windows Me Windows 2000 Windows XP

As you can see, various operating systems are supported by the International PGP site, which is working to establish PGP as a standard for encryption on the Internet. In addition to the standard PGP package, which provides for a number of applications, such as document encryption and email, a number of other products also are available, such as PGPdisk (for encrypting disks) and PGPphone (for making secure phone calls on the Internet).

The PGPi Project also is making PGP available in various languages, and also is currently translating the documentation. PGPi is a nonprofit organization dedicated to further developing and distributing PGP technology throughout the world. In addition, for some platforms, the source code is available so that you can examine it before compiling it on your system.