Many security-critical and widely deployed code bases are closed source, including some of the dominant operating system families for both servers and desktops. In order to assess the security of closed source software beyond the capabilities of fuzz-testing, binary auditing is a necessity.
In general, binary auditing is considered more difficult than auditing with source code. While this might seem like bad news for beginners, it could also be considered a benefit. There are far fewer people auditing binaries at this point in time, and fewer eyes make for easier work. Many bug classes that are virtually extinct in open source software still linger in closed source commercial code bases.
Binary auditing is still an imperfect science, and many things that can be quite easily verified while auditing source code are conversely quite difficult to determine while examining a binary. With practice and the help of some useful tools, much of the frustration associated with binary auditing can be removed.
Many security researchers do not stretch beyond the limitations of fuzz-testing when auditing commercial software. Although fuzz-testing has proven that it can reveal bugs in software, you really cannot fuzz all possible input patterns to any large piece of software in a reasonable amount of time. Binary auditing can offer a more complete view of the inner workings of an application and security flaws it might contain.