With the development of the Internet technology and the changes of the situation of Internet security, we witness exponential increase of the number of malicious software and their endless variants. Traditional detection methods cannot effectively and timely deal with such mass of malicious software data, making traditional anti-virus platform running on PC client cannot satisfy current security requirements any more, thus some major Internet security venders have been launching their ‘cloud security’ program. Under such background, it is urgent to develop some new effective and efficient techniques for malware detection. In this paper, we investigate malware detection techniques based on cloud computing, including mining massive software samples, and applying new clustering models/algorithms for event sequences into malware detection, to deal with the critical issues of malware as being of large amount, fast change, high-dimension and noise-laden. Furthermore, we propose a prototype of intelligent malware detection system for cloud security.