Fraud emails have become common problem in recent years. Fraud emails are a real threat to internet communication. In this paper, hybrid features are used for detecting fraud emails to determine how fast they have classified fraud emails and normal emails. Instead of hybrid feature, only content as a feature can also be used but most of the phishing email has similar content as normal email, so detection of phishing email is more complex and these approaches cannot give the higher rate classification. Hybrid feature selection approach based on combination of content based and header information.