Brute force attack
In cryptanalysis, a brute force attack is a method of defeating a cryptographic scheme by trying a large number of possibilities; for example, exhaustively working through all possible keys in order to decrypt a message. In most schemes, the theoretical possibility of a brute force attack is recognised, but it is set up in such a way that it would be computationally infeasible to carry out. Accordingly, one definition of "breaking" a cryptographic scheme is to find a method faster than a brute force attack.
The selection of an appropriate key length depends on the practical feasibility of performing a brute force attack.
Symmetric ciphers
For symmetric-key ciphers, a brute force attack typically means a brute-force search of the key space; that is, testing all possible keys in order to recover the plaintext used to produce a particular ciphertext.
In a brute force attack, the expected number of trials before the correct key is found is equal to half the size of the key space. For example, if there are 264 possible keys, a brute force attack would, on average, be expected to find a key after 263 trials.
For each trial of a candidate key the attacker needs to be able to recognise when he has found the correct key. The most straightforward way is to obtain a few corresponding plaintext and ciphertext pairs, that is, a known-plaintext attack. Alternatively, a ciphertext-only attack is possible by decrypting ciphertext using each candidate key, and testing the result for similarity to plaintext language — for example, English encoded in ASCII.
In general, a symmetric key cipher is considered secure if there is no method less expensive (in time, memory requirements, etc) than brute force; Claude Shannon used the term "work factor" for this. Nearly all ciphers lack a mathematical proof of security in this sense, although the one-time pad has been proven to provide perfect secrecy.
Symmetric ciphers with keys of length up to 64 bits have been broken by brute force attacks. DES, a widely-used block cipher which uses 56-bit keys, was broken by custom hardware in 1998 (see EFF DES cracker), and a message encrypted with RC5 using a 64-bit key was broken more recently by Distributed.net. In addition, it is commonly speculated that government intelligence agencies (such as the US NSA) can successfully attack a symmetric key cipher with long key lengths, such as a 64-bit key, using brute force. For applications requiring long term security, 128 bits is, as of 2004, currently thought a sufficient key length for new systems using symmetric key algorithms. NIST has recommended that 80-bit designs be phased out by 2015.
If keys are generated in a weak way, for example, derived from a guessable- password, it is possible to exhaustively search over a much smaller set, for example, keys generated from passwords in a dictionary. See password cracking and passphrase for more information.
Asymmetric algorithms
For asymmetric cryptography, a brute force attack usually involves tackling some difficult problem, such as large integer factorisation or calculating a discrete logarithm.
The situation with regard to asymmetric key algorithms is more complicated and depends on the individual encryption algorithm. Thus, the currently breakable key length for the RSA algorithm is at least 640 bits (i.e., it has been done publicly with the factorisation of RSA-640), and recent research developments suggest that 1024 bits might be breakable in the near to medium term future. For most elliptic curve asymmetric algorithms, the largest currently breakable key length is believed to be rather shorter, perhaps as little as 128 bits or so. A message encrypted with a 109 bit key by an elliptic curve encryption algorithm was publicly broken by brute force key search in early 2003. At this writing (as of 2004), 128 bit key lengths seem the minimum reasonable for elliptic curve algorithms, and 1024 bits for such asymmetric key algorithms as RSA.
Asymmetric algorithms have proven to be essential components of password-authenticated key agreement schemes that prevent brute-force attack on passwords and related data.
Theoretical limits
There is a physical argument that a 128 bit key is secure against brute force attack. It is argued that, by the laws of physics, in order to simply flip through the possible values for a 128-bit key (ignoring doing the actual computing to check it), one would need a device consuming at a minimum 10 gigawatts (about the equivalent of four large, dedicated nuclear reactors) running continuously for 100 years. The full actual computation—checking each key to see if you have found a solution—would consume many times this amount.
However, this argument assumes that the register values are changed using conventional set and clear operations which inevitably generate entropy. It has been shown that computational hardware can be designed not to encounter this theoretical obstruction: see reversible computing.
A more pragmatic approach to increasing the computational efficiency per unit of power utilized involves the recent trends toward multi-core processor technologies. Multi-core processors are able to boost computational performance for many types of computations while minimizing power utilization. (which would include the kind of computations required for cryptanalysis using brute force techniques) Though the current state of this technology is far from providing the reversible computatitive capability that would be required to make a brute force attack of a 128 bit key feasible, future innovations may significantly improve the estimated crack time and power requirements. This becomes more likely as multi-core processors are combined into innovative SMP (symetrically multi processor) architectures that by using well designed algorithms, optimize utiliziation of each core per unit of power across a massive array of processors.