Hacking AI: Exposing Vulnerabilities in Machine Learning