Free Tool Paints Picture Of Stealthy Attacks

The Honeynet Project has beefed up a free tool that helps spot attacks that can elude detection. The Picviz tool takes data from various log analysis sources and converts them into a multidimensional visual map of events.

Researchers have now added a graphical user interface to Picviz, which should make it easier to deploy and more attractive to a broader range of users. Picviz developers Sebastien Tricaud and Philippe Saade have published a paper (PDF) that details how Picviz works and how it gathers and renders data from traffic logs, database logs, SSH logs, syslogs, IPtables logs, Apache logs, and other sources.

Picviz's "parallel coordinates" approach represents an unlimited number of events in multiple dimensions, such as the protocol, URL, IP address, user agent, time frames, and other parameters. Parallel coordinates are multidimensional images used in aircraft collision-detection, as well as in other network tools. Picviz was developed to automate these images, according to Tricaud.

"Using parallel coordinates interactively makes finding issues rather natural and is very intuitive. Without a decent GUI, the program is only targeting a small set of people," says Tricaud, CTO of The Honeynet Project and chapter lead of the French Honeynet Project. The GUI and other features added to the tool make it more approachable to more users, he says.

Development of the tool first began a year ago, using mathematics to help provide a visual graphic of IP traffic that could help network managers and security professionals more quickly and efficiently pick up on anomalies or attacks. Picviz is different than existing visualization technologies, such as pie charts, histograms, or even 3D representations, because these technologies are restricted in the different dimensions of the data they can present, such as time, according to Honeynet Project members.

As part of Google's Summer of Code Program, Tricaud oversaw the work of student Victor Amaducci and Gabriel Cavalcante to add the new features to Picviz.

Picviz doesn't use signatures or statistical analysis to learn about an attack. "The starting point [with both of these approaches] is that you already know something about the attack," Tricaud says. Picviz's visual representation of the log data makes it easier to separate anomalies from regular traffic patterns, he says.

The tool also lets you interact with logs, he says. "You plot your data as a parallel coordinate plot, you play with them, you let your intuition rule the search, and, bingo: The important element in the data stands out in the foreground," he says. "You see things and their details very naturally. Another immediate benefit of this visual approach is that you can mix different types of dimensions -- time, numerical values, strings, etc. -- without loss of accuracy."

Picviz basically helps correlate data from multiple sensors and logs. "You always fight with the problem of having too many events coming at you. Then how do you get the information to quickly react? When the events arrive in a simple table, it is impossible to handle a large load," he says.

Have a comment on this story? Please click "Discuss" below. If you'd like to contact Dark Reading's editors directly, send us a message.