Cyber-security landscape has always been domi-nated by the constant arms race between security vendors and malware creators. With the increased usage of artificial intel-ligence in cyber-security products, malware writers had to use similar approaches to overcome various detection mechanisms. This paper focuses on using various adversarial machine learning algorithms to validate the effectiveness of machine learning models designed for malware detection, trained on 1 million binary files (both benign and malicious) with more than 24.000 features. We also evaluate a second approach that enhances training databases with samples obtained from generative methods in order to boost detection resilience.