In recent years, it has been observed that machine learning has started to exponentially grow in almost every domain. There is a very significant success rate in terms of its ability to deal with complex problems and coming up with solutions. But, recent studies have shown the existence of various vulnerabilities in various machine learning models and hence susceptible to attacks. In this paper, we will be highlighting certain major types of adversarial example attacks, backdoor attacks, poisoning approaches, and various defense techniques. There is also mention of biometric authentication system and the attacks which it may face. We will be systematically discussing the existing attacks on ML models and the defenses that may be taken. This paper contains a brief overview of research papers from the last five years. Last, future distractions for the network security are given for curious minds.