Previous research has shown that byte-level analysis of network traffic can be useful for network intrusion detection and traffic analysis. Such an approach does not require any knowledge of applications running on web servers or any pre-processing of incoming data. In this paper, we apply three n-gram techniques to the problem of HTTP attack detection. The goal is to provide a first line of defense by filtering the vast majority of benign HTTP traffic, leaving only a relatively small amount of suspect traffic for more costly processing. We analyze these n-gram techniques in terms of accuracy and performance. Our results show that we can attain equal or better detection rates at considerably less cost, in comparison to a previously developed HMM-based technique. We also apply these techniques to a highly realistic dataset consisting of four recent attacks and show that we obtain equally strong results in this case. Overall, these results indicate that this type of byte-level analysis is highly effective and practical.