An algorithm that successfully resisted cryptanalysis is considered strong. The strength criteria are usually formal and do not have any strict numerical expressions. For example, an algorithm that requires unattainable computing power will be considered as strong. Or the one that will require so much time to be hacked, that the encrypted information will lose its relevance and value to the cracker.

Strong systems are sub-divided into absolutely strong and computationally efficient. Absolutely strong are systems that generate a new secret key for each new message, the key’s length is greater or equal to the message itself, and the characters used in the key are random, independent, and their choice is equiprobable. An example of such a system was developed in 1917, the so-called one-time pad scheme, also known as the Vernam cipher. In contrast to the absolutely strong cryptographic systems the computationally efficient systems can have a number of vulnerabilities that in one way or another reduce the complexity of cracking.

The only way to guarantee the disclosure of any of the new algorithms (except for the absolutely efficient) is a brute force attack, in other words, a sequential scan of all possible keys. Computational complexity allows to evaluate this method in terms of required resources and the maximum time needed for scanning, which eventually will give the performance rate and the potential financial cost.

Besides the direct key scanning that leads to a further analysis, there can be other types of attacks, such as differential and linear cryptanalysis – these methods help to find the vulnerabilities of an algorithm and decrease the cryptographic resistance of a system. For example, for many symmetric encryption systems there are so-called weak keys. In addition to mathematical analysis there can be some other methods for example, the attacks on the implementation or on side-channel cryptanalysis.

As long as the encryption algorithm is not compromised, it is considered efficient. Longer the algorithm is published and is in the public domain, the greater is the number of experts working on deciphering it and more reliable its efficiency is. It is considered that if the algorithm doesn’t reveal any serious vulnerability for five years, it can be used to protect sensitive information.