Software Reliability

Between system security measures (security kernels, access control measures, strong cryptography, etc.) and good network security measures (firewalls, intrusion detection systems, auditing mechanisms), it seems as if computer security is pretty much done. Why then, are computers and networks so insecure? Why are we seeing more computer security vulnerabilities in the media, and not less? Why aren't things getting better?

The problem is that security measures such as cryptography, secure kernels, firewalls, and everything else work much better in theory than they do in practice. In other words: Security flaws in the implementation are much more common, and much more serious, than security flaws in the design. So far, Part 2 has talked about design. This chapter is about implementation.


In June 1996, the European Space Agency's Ariane 5 rocket exploded after launch because of a software error: The program tried to stick a 64-bit number into a 16-bit space, causing an overflow. Its lessons are particularly relevant to computer security.

Basically, there was a piece of code written for the Ariane 4 rocket that dealt with the rocket's sideways velocity. At 36.7 seconds after launch, the guidance system's computer tried to convert this velocity measurement from a 64-bit format to a 16-bit format. The number was too big, which caused an error. Normally, there would be extra code that watches for these sorts of errors and recovers gracefully. But the ...

Get Secrets and Lies: Digital Security in a Networked World, 15th Anniversary Edition now with the O’Reilly learning platform.

O’Reilly members experience live online training, plus books, videos, and digital content from nearly 200 publishers.