Archive for the ‘papers’ Category

Eudaemon: Argos-like protection on the desktop

Tuesday, May 6th, 2008

“Essentially, Eudaemon is a ‘good spirit’ that possesses processes in order to protect them form evil.”

Despite all of the attractive properties of Argos and the various signature generators, we are still missing some attacks: those on the client-side. Honeypots have as a disadvantage that they mostly act like servers and wait for attackers to contact them. However, client-side attacks are on the rise. In such attacks, the client visits a malicious website and is infected by the content. Honeypots never see such attacks.

Moreover, the Argos honeypot only protects the exact configuration that it is actually running on top of its emulator. If the production machines have a slightly different configuration (e.g., a different version of the OS, applications, or plugins), Argos may miss attacks. So, ideally we would like to protect the `real’ machines.

In the Eudaemon project, some of the groundwork is done for implementing techniques to bring Argos-like honeypot technology to the desktop of normal users. In other words: client-side protection. This is a major challenge as the emulator that implements all the required instrumentation to detect attacks incurs a slowdown of a factor 15-20. As a result, running a desktop PC in instrumented mode constantly is probably not acceptable.

Instead, the idea is to take over running processes and force them to continue running in honeypot mode when needed or when possible. For instance, in the former case one may run a browser in honeypot mode when we click on an unknown URL in an email message. The browser is then protected against browser attacks by malicious server content. In the latter case, one may switch all of the user’s networking applications to honeypot mode when the machine is idle. For instance, we may switch to honeypot mode as a screensaver.

The mechanism for doing this is described in the Eudaemon paper presented at EUROSYS 2008 in Glasgow. [bibtex]

WIT: Write Integrity Testing

Wednesday, April 23rd, 2008

Cool stuff from the people at MSR Cambridge. Their latest Oakland paper on write integrity checking provides a nice solution for various memory exploits that may otherwise lead to a machine being compromised. Their method uses point-to-analysis at compile time to generate the control flow graph as well as the set of objects that can be written by each instruction. It then instruments these instructions to make sure that they do not write into with objects that are not in the set (and also to ensure that indirect control flow transfers are valid). Runtime overhead is very small. I guess the main catch is that you have to sit on the source code – which is not the case for anyone other than Microsoft. Nevertheless, very good work.

Temporal Search: Detecting Hidden Malware Timebombs with Virtual Machines

Friday, March 14th, 2008

Imagine that a new worm is released at 10pm and that after half an hour has infected a few thousands of computers (10.30pm). Suppose that the goal of this worm is to coordinate a DDoS attack against a critical information system, such as root DNS servers, at 11.30pm. Suppose also that we can avert this attack if we knew about it half an hour before. Are 30 minutes enough for humans to analyze the worm and find out its malicious goal? I believe not.

So we need a way to automatically analyze the above worm, find the time that this attack will take place and the details of the attack (DDoS against root DNS servers). Crandall et al. propose a framework to automate the process of finding the time that the attack will take place. Although manual analysis is required for finding out the details of the attack, this is a step to the right direction.

To be more specific, the authors propose a novel virtual-machine based analysis technique for automating the process of discovering the timetable of a piece of malware. The obvious advantage of automating this process is that it requires less time than careful human analysis. Given that malware is expected to spread faster and faster in the years to come, such information provide invaluable assistance defending against malware.

The main contributions of this paper is that it proposes a framework for a problem space nobody has looked at the past and that it identifies many of the challenges in this new problem space.

Source: Temporal Search: Detecting Hidden Malware Timebombs with Virtual Machines, Jedidiah R. Crandall, Gary Wassermann, Daniela A. S. de Oliveira, Zhendong Su, S. Felix Wu, and Frederic T. Chong.