Basic genetic-algorithm-neural-network (GANN) pattern with a self-organizing security example

The anti-system adversarial community is characterized as a self-organizing system-of-systems, noted collectively for its leadership in rapid evolution and innovative advancement; widening the gap between security cost and security losses. It appears that system security strategy cannot hope to even achieve parity without a comparable self-organizing strategy. Toward that end a project is underway to catalog re-usable patterns of self-organizing security of many kinds, principally found in natural systems, but also seen in recent computational approaches. One class of pattern of special interest involves discovery of previously unseen threats and attacks. In general this class of pattern has aspects of learning, innovation, and evolution as capability objectives. The genetic algorithm is one such pattern. Another such pattern is seen in artificial neural networks. Combining the two into a Genetic Algorithm augmented Neural Network, often called GANN, has considerable recent history in the literature. Not many of these are directly related to security applications. Some security-application work shows GANNs employed for feature selection provide enhanced learning performance and accuracy, and avoidance of local minimum traps. This paper adds the GANN pattern to the self-organizing security pattern catalog, and applies the pattern to a self-organizing security application under development.