Adversarial Learning

Machine learning techniques have been applied to wide range of applications due to satisfying performance. Many applications, for example spam filtering, virus detection and terrorism detection, may involve an adversary who misleads the system on purpose by manipulating the data. As standard machine learning methods assume the distributions of training and unseen samples are the similar (implicitly), these methods may be suffered from adversarial attack and their performance will drop significantly. Adversarial Learning addresses the issues including vulnerability identification, attack strategy and coutermeasure algorithm design.