Opportunities and Risks of Volunteer Computing
From Asia at Home
Researchers agree: volunteer computing, volunteer thinking and volunteer teaching offer great possibilities for resource sharing and collective knowledge building. But what about the risks users take when installing the software and running client applications? David Anderson, inventor of BOINC (http://boinc.berkeley.edu) and researcher at the University of California, points out two risks and how they are being addressed.
First of all, a hacker could potentially replace an application with malware, with bad consequences for all the volunteers using the application. This can be prevented by digitally “signing” applications on a physically isolated computer. Second, application might be valid but buggy. “An account-based sandbox is the answer here”, explains David Anderson.
Projects also need to be protected from malicious volunteers. For example, volunteers might upload bad data, falsifying results. This can be prevented by sending the same job the two volunteers and then cross-checking the results obtained. The replication can be skipped when a volunteer has gained a sufficiently good reputation. Some projects might involve sensitive data – for example, gene sequences. Such data cannot be hidden from volunteers, even if files are encrypted. However, it may be possible to divide a data set into small pieces that are not sensitive in isolation.
Last but not least, there is an organisational issue to be dealt with: creating a volunteer computing project has a start costs and requires a diverse set of skills, which limits its use by individual scientists and research groups.
Article by V. Krebs, http://www.icvolunteers.org
