Homepage

Description

This file contains 9 sets of sanitized user data drawn from the command histories of 8 UNIX computer users at Purdue over the course of up to 2 years (USER0 and USER1 were generated by the same person, working on different platforms and different projects). The data is drawn from tcsh(1) history files and has been parsed and sanitized to remove filenames, user names, directory structures, web addresses, host names, and other possibly identifying items. Command names, flags, and shell metacharacters have been preserved. Additionally, **SOF** and **EOF** tokens have been inserted at the start and end of shell sessions, respectively. Sessions are concatenated by date order and tokens appear in the order issued within the shell session, but no timestamps are included in this data. For example, the two sessions: # Start session 1 cd ~/private/docs ls -laF | more cat foo.txt bar.txt zorch.txt > somewhere exit # End session 1 # Start session 2 cd ~/games/ xquake & fg vi scores.txt mailx john_doe '@' somewhere.com exit # End session 2 would be represented by the token stream **SOF** cd <1> # one "file name" argument ls -laF | more cat <3> # three "file" arguments > <1> exit **EOF** **SOF** cd <1> xquake & fg vi <1> mailx <1> exit **EOF**

Related Papers

• Stefan Aeberhard and Danny Coomans and De Vel. THE PERFORMANCE OF STATISTICAL PATTERN RECOGNITION METHODS IN HIGH DIMENSIONAL SETTINGS. James Cook University. [link]