Hands-On Artificial Intelligence for Cybersecurity

Getting to know GANs

We have said that the intuition on which GANs are based entails putting two NNs in competition with one another in order to improve the overall results. The term adversarial refers specifically to the fact that the two NNs compete between themselves in completing their respective tasks. The outcome of this competition is an overall result that cannot be further improved, thereby attaining an equilibrium condition.

A typical example of using GANs is the implementation of a particular NN, called a generative network, with the task of creating an artificial image that simulates the characteristics of a real image. A second NN, called the discriminator network, is placed in competition with the first one (the generator) in ...