Provable security refers to any type or level of computer security that can be proved. It is used in different ways by different fields.
Usually, this refers to mathematical proofs, which are common in cryptography. In such a proof, the capabilities of the attacker are defined by an adversarial model (also referred to as attacker model): the aim of the proof is to show that the attacker must solve the underlying hard problem in order to break the security of the modelled system. Such a proof generally does not consider side-channel attacks or other implementation-specific attacks, because they are usually impossible to model without implementing the system (and thus, the proof only applies to this implementation).
Outside of cryptography, the term is often used in conjunction with secure coding and security by design, both of which can rely on proofs to show the security of a particular approach. As with the cryptographic setting, this involves an attacker model and a model of the system. For example, code can be verified to match the intended functionality, described by a model: this can be done through static checking. These techniques are sometimes used for evaluating products (see Common Criteria): the security here depends not only on the correctness of the attacker model, but also on the model of the code.
Finally, the term provable security is sometimes used by sellers of security software that are attempting to sell security products like firewalls, antivirus software and intrusion detection systems. As these products are typically not subject to scrutiny, many security researchers consider this type of claim to be selling snakeoil.
In cryptography
In cryptography, a system has provable security if its security requirements can be stated formally in an adversarial model, as opposed to heuristically, with clear assumptions that the adversary has access to the system as well as enough computational resources. The proof of security (called a "reduction") is that these security requirements are met provided the assumptions about the adversary's access to the system are satisfied and some clearly stated assumptions about the hardness of certain computational tasks hold. An early example of such requirements and proof was given by Goldwasser and Micali for semantic security and the construction based on the quadratic residuosity problem. Some proofs of security are in given theoretical models such as the random oracle model, where real cryptographic hash functions are represented by an idealization.
There are several lines of research in provable security. One is to establish the "correct" definition of security for a given, intuitively understood task. Another is to suggest constructions and proofs based on general assumptions as much as possible, for instance the existence of a one-way function. A major open problem is to establish such proofs based on , since the existence of one-way functions is not known to follow from the conjecture.
Controversies
Several researchers have found mathematical fallacies in proofs that had been used to make claims about the security of important protocols. In the following partial list of such researchers, their names are followed by first a reference to the original paper with the purported proof and then a reference to the paper in which the researchers reported on flaws: V. Shoup; A. J. Menezes; A. Jha and M. Nandi; D. Galindo; T. Iwata, K. Ohashi, and K. Minematsu; M. Nandi; J.-S. Coron and D. Naccache; D. Chakraborty, V. Hernández-Jiménez, and P. Sarkar; P. Gaži and U. Maurer; S. A. Kakvi and E. Kiltz; and T. Holenstein, R. Künzler, and S. Tessaro.