Popularity of Android smart phone has led to exponential increase of sophisticated malware threats prompting the academia research, security researchers and Anti Virus (AV) industry to look for smart detection methods to protect user against malware app threat. Statistical signature methods play a vital role to stop the malware authors spreading malicious content through apps. In this research, we present DroidOLytics, a statistical signature approach that creates improbable feature signature to detect unseen malicious apps from third party and official android market. Statistical signature is robust against repackaged and code obfuscated malware, popular app obfuscation techniques. DroidOLytics is a syntactic approach that finds regions of statistical similarity with known malware to detect variants of known malware families.