There is a name for this in machine learning, it's called a 'poisoning attack'. Essentially you add false data into a training set. An example: https://arxiv.org/abs/1703.01340