Security Researchers: How do you choose what services or software you are going to audit next?
I am referring to finding vulnerabilities and developing exploits in various applications. I am wandering how decide what software you will target next?
By scanning the internet, the Nmap team has elaborated a list of the most commonly exposed TCP and UDP services. You can make use of this list in order to determine the next service you will be auditing next, assuming that you want to target the most commonly used ones.
I audit a lot of code, I write exploits, and I have accumulated more than 50 CVE's over the course of about 6 years of bug hunting.
When I went on my first serious bug hunt I was looking for weak projects that had not be extensively audited by the community(or milw0rm back in the day). To do this I used SourceForge's Advanced Search, which has changed a lot. Basically I was looking for a PHP project that people where downloading and using, but wasn't very popular. Lets say around ~1,000 downloads and less than 1 year old. I found Ultimate PHP Board which turned out to be very insecure.
As time progressed I got bored with insecure projects so I changed my tactics entirely and started going after popular projects. For example I exploit PHPMyAdmin which is the most downloaded PHP application.
After years of penetration testing and application development you get a kind of 6th sense into how the code works and where the problem areas can be. You can look at a piece of functionality and write an implementation in your head and pick out where things could go wrong. So on a penetration test I always ask my self the same question: "What is the worst that could happen?". And then go out and focus my testing based on this question.
I'll give you a good example. I saw an advertizement for Canonical Landscape (Maker of Ubuntu). In the ad it showed a feature of Landscape where you could execute a command on every machine you own as root. I thought to my self, "What if it was vulnerable to CSRF?". I signed up for a free trail, and sure enough, you could gain remote root on every machine using a single forged HTTP request. Outch! (I also got remote root on cPanel with CSRF, and I earned a severity metric :)