Bandwidth demands of communication networks are rising permanently. Thus, the requirements to modern routers regarding packet classification are rising accordingly. Conventional algorithms for packet classification use either a huge amount of memory or have high computational demands to perform the task. Using a hash function in order to classify packets is promising regarding both memory and computation time. However, such a hash function needs to be of high performance and cheap in hardware costs. These two design goals are contradictory. To limit the costs of a hardware implementation, known good hash functions, as used for software implementations of encryption algorithms, are applicable to only a limited extend. To achieve the goals mentioned above, an adaptive hash function is needed. In this paper, an approach for a hardware packet classifier using an evolvable hash function is presented. It consists of an evolutionary algorithm which is entirely implemented in hardware.