In this paper, we have identified malicious web servers with our high interaction client honeypot Capture-HPC. As part of future work, we would like to identify malicious web servers with our low interaction client honeypot HoneyC and compare the results. We suspect that this comparison will give us insights into the detection accuracy, in particular false negatives, of each client honeypot technology.

Further, we would like to expand our research to client-side attacks that target browser plug-ins as well as non-browser client applications. The data we have collected as part of this study already shows that browser plug-ins, such as QuickTime and Winzip, are targeted. A closer look at browser plug-ins will allow us to assess the magnitude of the problem. In addition to browser plug-ins, we would like to evaluate the risk to non-browser applications, such as Microsoft Office, Adobe Acrobat Reader, etc. Many remote execution vulnerabilities have been publicly disclosed for these client applications and it is suspected that they are also targeted. Our future research will determine the extent of the threat.

In addition, time sensitive behavior was not addressed by our study extensively. While we observed that malicious URLs tend to stop soliciting malicious behavior after some time has passed, a representative model of the disappearance and appearance would be necessary in order to assess growth rates of client side attacks. A trend analysis would be required. Along these lines, we would also like to asses how quickly new exploits appear on the Internet. Interesting time factors to consider are the disclosure of the vulnerability, public availability of the exploit and the availability of the patch.