Hello,
What Dr. Jakobsson has described is a reputation system.
At
Virus Bulletin 2009, Symantec gave a presentation on reputation systems: "
Using the wisdom of crowds to address the malware long tail ," which cited data from one that began development in 2006. While I do not claim to understand the system, in a nutshell, it seems to work by generating a hash for files after they are downloaded or when they are to be executed, and sends this to Symantec along with some metadata, such as source IP/host, filename, path specification on the local host, date and time stamp on the file and other useful information, which is sent to Symantec, initially to provide a quick lookup, but more information can be sent if additional analysis is required. Symantec's client software can then display a message saying "Program
XYZ.EXE has been seen
n time(s) over the course of
n day(s)/week(s)/month(s)." along with some suggestions about how safe it is likely to be based on new/unique program files more likely to be untrusted (higher potential for malcode) and older, commonly program files having a higher degree of trust.
One advantage of this approach is that it quickly allows malcious files encoded using server-side polymorphism to be quickly identified, as well as the sites hosting them. This negates the technique used by the bad guys to constantly modify code to in order to escape detection by anti-virus software.
Regards,
Aryeh Goretsky