The voluminous malware variants in the cyberspace are responsible for a variety of online criminal activities. In this talk, I will present a big data approach to fighting malware. First, I will discuss how to develop a Neyman-Pearson classifier for classifying malware variants into their corresponding families with bounded false positive rates. Second, I will talk about how to label malware samples from AV (Anti-Virus) scanners' detection results among which there may exist conflicts.