and broken for real, yet continued use will endanger all messages so "protected." Thus, it is a very reasonable policy to not adopt a widely-used cipher, and to change ciphers periodically. Cryptology The field of study which generally includes steganography, cryptography and cryptanalysis. Current The measure of electron flow, in amperes. Current is analogous to the amount of water flow, as opposed to pressure or voltage. A flowing electrical current will create a magnetic field around the conductor. A changing electrical current may create an electromagnetic field. dB decibel. DC Direct Current: Electrical power which flows in one direction, more or less constantly. As opposed to AC. Most electronic devices require DC at least internally for proper operation, so a substantial part of modern design is the "power supply" which converts 120 VAC wall power into 12 VDC, 5 VDC and/or 3 VDC as needed by the circuit and active devices. Debug The interactive analytical process of correcting the design of a complex system. A normal part of the development process, although when bugs are not caught during development, they can remain in production systems. Contrary to naive expectations, a complex system almost never performs as desired when first realized. Both hardware and software system design environments generally deal with systems which are not working. (When a system really works, the design and development process is generally over.) Debugging involves identifying problems, analyzing the source of those problems, then changing the construction to fix the problem. (Hopefully, the fix will not itself create new problems.) This form of interactive analysis can be especially difficult because the realized design may not actually be what is described in the schematics, flow-charts, or other working documents: To some extent the real system is unknown. When a system has many problems, the problems tend to interact, which can make the identification of a particular cause very difficult. This can be managed by "shrinking" the system: first by partitioning the design into components and testing those components, and then by temporarily disabling or removing sections so as to identify the section in which the problem lies. Eventually, with enough testing, partitioning and analysis, the source of any problem can be identified. Some "problems," however, turn out to be the unexpected implications of a complex design and are sometimes accepted as "features" rather than the alternative of a complete design overhaul. Decipher The process which can reveal the information or plaintext hidden in message ciphertext (provided it is the correct process, with the proper key). The inverse of encipher. Decryption The general term for extracting information which was hidden by encryption. Deductive Reasoning In the study of logic, reasoning about a particular case from one or more general statements; a proof. Also see: inductive reasoning and fallacy. Defined Plaintext Attack A form of attack in which the Opponent can present arbitrary plaintext to be enciphered, and then capture the resulting ciphertext. The ultimate form of known plaintext attack. A defined plaintext attack can be a problem for systems which allow unauthorized users to present arbitrary messages for ciphering. Such attack can be made difficult by allowing only authorized users to encipher data, by allowing only a few messages to be enciphered between key changes, by changing keys frequently, and by enciphering each message in a different random message key. Degrees of Freedom In statistics, the number of completely independent values in a sample. The number of sampled values or observations or bins, less the number of defined or freedom-limiting relationships or "constraints" between those values. If we choose two values completely independently, we have a DF of 2. But if we must choose two values such that the second is twice the first, we can choose only the first value independently. Imposing a relationship on one of the sampled value means that we will have a DF of one less than the number of samples, even though we may end up with apparently similar sample values. In a typical goodness of fit test such as chi-square, the reference distribution (the expected counts) is normalized to give the same number of counts as the experiment. This is a constraint, so if we have N bins, we will have a DF of N - 1. DES The particular block cipher which is the U.S. Data Encryption Standard. A 64-bit block cipher with a 56-bit key organized as 16 rounds of operations. Decibel Ten times the base-10 logarithm of the ratio of two power values. Denoted by dB. One-tenth of a bel. When voltages or currents are measured, power changes as the square of these values, so a decibel is twenty times the base-10 logarithm of the ratio of two voltages or currents. Decimal Base 10: The numerical representation in which each digit has an alphabet of ten symbols, usually 0 through 9. Also see: binary, octal, and hexadecimal. Design Strength The keyspace; the effort required for a brute force attack. Deterministic A process whose sequence of operations is fully determined by its initial state. A mechanical or clockwork-like process whose outcome is inevitable, given its initial setting. Pseudorandom. Dictionary Attack Typically an attack on a secret password. A dictionary of common passwords is developed, and a brute force attack conducted on the target with each common password. Differential Cryptanalysis A form of attack in which the difference between values (or keys) is used to gain some information about the system. Also see Differential Cryptanalysis: A Literature Survey, in the Literature Surveys and Reviews section of the Ciphers By Ritter page. Diffusion Diffusion is the property of an operation such that changing one bit (or byte) of the input will change adjacent or near-by bits (or bytes) after the operation. In a block cipher, diffusion propagates bit-changes from one part of a block to other parts of the block. Diffusion requires mixing, and the step-by-step process of increasing diffusion is described as avalanche. Diffusion is in contrast to confusion. Normally we speak of data diffusion, in which changing a tiny part of the plaintext data may affect the whole ciphertext. But we can also speak of key diffusion, in which changing even a tiny part of the key should change each bit in the ciphertext with probability 0.5. Perhaps the best diffusing component is substitution, but this diffuses only within a single substituted value. Substitution-permutation ciphers get around this by moving the bits of each substituted element to other elements, substituting again, and repeating. But this only provides guaranteed diffusion if particular substitution tables are constructed. Another alternative is to use some sort of Balanced Block Mixing which has an inherently guaranteed diffusion, or a Variable Size Block Cipher construction. Also see Overall Diffusion. Digital Pertaining to discrete or distinct finite values. As opposed to analog or continuous quantities. Diode An electronic device with two terminals which allows current to flow in only one direction. Distribution In statistics, the range of values which a random variable, and the probability that each value or range of values will occur. Also the probability of test statistic values for the case "nothing unusual found," which is the null hypothesis. If we have a discrete distribution, with a finite number of possible result values, we can speak of "frequency" and "probability" distributions: The "frequency distribution" is the expected number of occurrences for each possible value, in a particular sample size. The "probability distribution" is the probability of getting each value, normalized to a probability of 1.0 over the sum of all possible values. Here is a graph of a typical "discrete probability distribution" or "discrete probability density function," which displays the probability of getting a particular statistic value for the case "nothing unusual found": 0.1| *** | * * Y = Probability of X Y | ** ** y = P(x) | **** **** 0.0 X Unfortunately, it is not really possible to think in the same way about continuous distributions: Since continuous distributions have an infinite number of possible values, the probability of getting any particular value is zero. For continuous distributions, we instead talk about the probability of getting a value in some subrange of the overall distribution. We are often concerned with the probability of getting a particular value or below, or the probability of a particular value or above. Here is a graph of the related "cumulative probability distribution" or "cumulative distribution function" (c.d.f.) for the case "nothing unusual found": 1.0| ****** | ** Y = Probability (0.0 to 1.0) of finding Y | * a value which is x or less | ** 0.0 -****** X The c.d.f. is just the sum of all probabilities for a given value or less. This is the usual sort of function used to interpret a statistic: Given some result, we can look up the probability of a lesser value (normally called p) or a greater value (called q = 1.0 - p). Usually, a test statistic is designed so that extreme values are not likely to occur by chance in the case "nothing unusual found" which is the null hypothesis . interact, which can make the identification of a particular cause very difficult. This can be managed by "shrinking" the system: first by partitioning the design into components and testing. after the operation. In a block cipher, diffusion propagates bit-changes from one part of a block to other parts of the block. Diffusion requires mixing, and the step-by-step process of increasing. diffusion, in which changing a tiny part of the plaintext data may affect the whole ciphertext. But we can also speak of key diffusion, in which changing even a tiny part of the key should change