On Sep 15, 3:20pm, Jamie Lawrence wrote:
Also, I do disagree with your statement "security through obscurity is no security at all." A rather high degree of security can be had through obscurity, but it is often entirely unpredictable whether or not a particlar 'obscurity method' will be secure or not (any 15 year old hiding cigarettes under the bed can attest to that).
This is absolutely correct. Keeping your secret key a secret _is_ security by obscurity, although in a much wider context than most people would use the term. In addition, it is also particularly effect if what is being obscured is sufficiently secure already, as it just adds another layer of protection. For example, if I decided to superencrypt using some publicly known and reasonably trusted ciphers (let's say DES, LOKI and IDEA), and decided to keep the algorithms I had used and the order I had used them a secret, I have _not_ decreased my security. The obscurity does not deduct from the security of these already moderately trusted ciphers, and the work which would have to be added to figure out what I have done increases it (although by an amount which is probably arguable). I am sure that this is a point almost everyone here understands this concept, but it's amazing how many times the argument "it's a secret, therefore it's insecure" comes out. It's only really insecure if the thing you're keeping a secret is, and even then you have not decreased it's effective security by obscuring it. The TLA's understand this concept well, which is one of the reasons they classify almost everything they do. One non-obvious fact is that in the environment most governments use crypto (eg. widely distributed sites with key distribution channels which are more easily compromised than the crypto hardware), that the design of the cipher may be easier to keep secret than the key itself. As such, the use of security by obscurity in the design of the cipher itself is a lot more effective than most people would give it credit for. Ian.