Evolutionary design of domain specific non-cryptographic hash functions

Hash functions are inseparable part of modern computer world. Fast associative arrays so popular among computer programmers for their robustness and simplicity, are based on them. Their performance greatly depends on their design and although their roots are deep in the past, the topic of designing a well performing hash function is still often discussed today. There is currently a plenty of exceptional implementations of generic hash functions and their numbers are rising. Such functions are not constrained to a concrete set of inputs, they perform on any input. On the other hand, there are use cases when input domain is known in advance. In such case, there is a room for an improvement by designing a specific hash function thus reaching better level of performance in comparison with a generic hash function. However designing a hash function is not a trivial task. There are no rules, standards or guides. In case of manual design the hash function author has to rely on his/her knowledge, experience, inventiveness and intuition. This fact opens up a space for different techniques such as evolutionary algorithms, an unconventional approach to solve certain problem inspired by the process of species reproduction. In this paper hash functions are designed for the domain of IP addresses using genetic programming. Genetic programming algorithm parameters are accurately chosen so the evolved functions will perform on its best. The experiments proved that the developed hash functions are better in comparison with generic hash functions in terms of both the speed and the performance.