Thursday, November 6, 2008

Nmap's Fourth GSoC: Success Stories and Lessons Learned

By Gordon "Fyodor" Lyon, author of the Nmap Project and GSoC Mentor

The Nmap Security Scanner Project was honored to participate in our fourth Google Summer of Code(tm)! The pencils-down date was two months ago, but so much code was produced that we're just now finishing the integration process. I finally have time to reflect on these last four years, what GSoC has brought us, and the lessons it has taught us.

In 2005 (detailed writeup), 70% (7 out of 10) students succeeded, and they tackled some wonderful projects! This year we begin work on our new Zenmap GUI (then named Umit), Ncat network communication utility, and 2nd generation OS detection system. Doug Hoyte first made major contributions that summer, and continues helping to this day. I was the mentor for all 10 students, and I had them all send me patches rather than providing SVN access. Nmap didn't even have a public SVN tree back then.

In 2006 (full writeup), I had a better idea of what works and what doesn't and was able to improve the success rate to 80% (8 out of 10). Perhaps the most exciting project was the Nmap Scripting Engine (NSE), which has become one of Nmap's most compelling features. It allows users to write (and share) simple scripts to automate a wide variety of networking tasks. We also finished and integrated the 2nd generation OS detection system, and Zenmap (Umit) continued to improve. I again mentored the students myself without providing SVN access.

In 2007 (full writeup), our success rate grew again to 83% (5 of 6)! I attribute part of the success to me being less of a control freak. For example, I took only 4 students compared to 10 the previous year. The remaining two 2006 students were mentored by Diman Todorov, who created NSE as a 2006 SoC student. I also made the Nmap SVN server public and provided commit access to the students. This year we formally integrated Zenmap into the Nmap build system and packages, making massive improvements along the way. This Summer also introduced David Fifield to the Nmap project and was the first SoC for Kris Katterjohn. Both of them have been prolific developers ever since then.

Enough with the history—let's take a look at our 2008 results! I'm happy to report that we had an 86% (6 out of 7) success rate. In other words, our success rate has increased every single year! I like to credit improved processes and interaction based on what we've learned before, but it also helps that we invite the best students back in later years. We've never had a 2nd year (or more) student fail. This year we expanded to three mentors, all of whom (except for me) were former SoC students. Now let's look in detail at our 2008 SoC accomplishments:

  • Vladimir Mitrovic spent the summer improving the Zenmap GUI, under
  • David Fifield's expert mentorship. They made huge usability and stability improvements, but the pinnacle of their summer achievement was clearly the scan aggregation and topology features! Scan aggregation allows you to conduct multiple scans at different times and add them seamlessly to your existing results. Topology draws a beautiful interactive diagram like this of the discovered network:


  • Jurand Nogiec also worked with David on Zenmap, and was responsible for many key UI improvements which now seem obvious in hindsight. For example, he added a cancel button for aborting a scan in progress without clearing the Nmap output, and he added context-sensitive help to the many dozens of options in the Profile Editor. He also made numerous improvements to the command entry field for people who like to type Nmap command directly, while still benefiting from Zenmap's visual and searchable presentation of results.


  • Patrick Donnelly made substantial NSE infrastructure improvements. He added mutex support and an NSE Standard Library, fixed some serious bugs, and rewrote and optimized a substantial amount of code (particularly the nse_init system). But his crowning accomplishment was the NSEDoc system, which uses special comments and variables in script and library code to generate a comprehensive documentation portal.


  • Kris Katterjohn, who already had hundreds of useful Nmap patches to his name, returned for 2008 to write hundreds more! There is no way I can list everything he did here, particularly as his contributions ranged all over the map from writing NSE libraries (such as the username/password module and the standardized communication library) to improving Windows support (adding IPv6 and OpenSSL). His biggest project has been finishing up Ncat, our advanced Netcat replacement (which began as a 2005 SoC project by Chris Gibson). Ncat is now integrated with Nmap in our latest SVN revision.
  • Michael Pattrick was David's third student, and he accomplished a wide variety of tasks. For example, he created a new OSAssist application for testing and integrating the thousands of Nmap OS detection submissions sent in by Nmap users all over the world. With OSAssist, integration is more accurate and much less tedious. Michael also built two prototypes (one in Perl and then another in C++) for an Ndiff application which compares two or more scan output files and prints out any changes. The prototypes proved so popular that David wrote a final version in Python which is now integrated with Nmap in our latest SVN revision.
  • Philip Pickering spent the summer working on NSE scripts and libraries. We've already incorporated his libraries for binary data manipulation, DNS queries, Base64 encoding, SNMP, POP3, and cryptographic hashes. We've also incorporated several scripts he wrote utilizing these new libraries.


In addition to these core Nmap projects, 5 students were sponsored to work on the UMIT Nmap GUI (now a separate project led by Adriano Marques). Four of their five students passed, as described here.

Please join me in congratulating all these students for their excellent work! I'm particularly pleased that many of the SoC students have continued contributing even though the summer has ended. I'm looking forward to GSoC 2009 (assuming it is held again and they invite us), but 2008 will be a tough year to top!

Link - from Google Open Source Blog
Related From Google Blogs:
Gmail for mobile: faster, smoother, and now in more languages
More changes to Gmail contact manager
SAS 70 Type II for Google Apps
Ending our agreement with Yahoo!

No comments: