Which of the following types of attacks would a malicious actor MOST likely use to target an AI model's training data and compromise the accuracy of the model's...