Not a professional reverse engineer. But the jar is not obfuscated and is fairly easy to understand.
The jar loads a "model" which is a serialised version of a class containing keywords from a url constructed from the applet params "codebase" and "getmodelurl".
It looks into
1. Chrome, firefox, and IE histories
Matches history with sites in the "model"
2. Main filesystem
It enumerates all the files in the file system, checking if it is either a picture, or a movie, then checks if the filename matches the list of keywords obtained from the "model".
