The Code Book - S. Singh

One of the very few examples I encountered - but to be honest, I tend to avoid the category as much as possible - of a good divulgation book about a technical subject. The only reason I read it is because I received at as a present.

Anyway, it’s a very good read which I think (among other things) makes one point very clear - there are two questions, 1. whether a message can be decyphered knowing the technique with which it has been encrypted, 2. whether a message can be decyphered in a reasonable amount of time, and the first question is the one mathematicians like, the second one is the one which needs an answer. There are abundant cases of coded message having no particularly spectacular cyphering technique, very involved but nothing new, and it took forever to crack the cypher just because nobody really knew which technique was used to begin with - are letters being substituted or vowels, or both, or maybe sometimes even words, and are they substituted with one number, two, three, or a mixture, or are maybe sounds being swapped instead, etc. It is similar to the distinction between parameters and hyperparameters. It seems that the focus, scientifically speaking, is on whether it is possible to determine the parameters given the hyperparameters, but the hyperparameters are seldomly known, and, again, time is a key component.

The book making this (admittedly saddening) point clear is a testament to the transparency and honesty the argument is dealt with. My takeaway is that if you really really really want to keep something secret you should just change the hyperparameters as often as possible, since frequency is key to all kind of decodifications, and otherwise one should make peace with other people being able to hear what you say.