The increasing volume and variety of malware is posing a serious security threat to the Internet today and is one of the main apprehensions for the security community for the last few years. The traditional security systems like Intrusion Detection System/Intrusion Prevention System and Anti-Virus (AV) software are not able to detect unknown malware as they use signature based methods. In order to solve this issue, static and dynamic malware analysis is being used along with machine learning algorithms for malware detection and classification. The main problems with these systems is that they have high false positive and false negative rate and the process of building classification model takes time (due to large feature set) which hinders the early detection of malware. Thus, the challenge is to select a relevant set of features, so that, the classification model can be built in less time with high accuracy. In this paper, we present a system that addresses both the issues mentioned above. It uses an integration of both static and dynamic analysis features of malware binaries incorporated with machine learning process for detecting zero-day malware. The proposed model is tested and validated on a real-world corpus of malicious samples. The results show that the static and dynamic features considered together provide high accuracy for distinguishing malware binaries from clean ones and the relevant feature selection process can improve the model building time without compromising the accuracy of malware detection system.