Cryptographers design and implement cryptographic protocols. This includes management of crypto keys. Cryptanalysts try to break such crypto systems, and so defeat the goals of the cryptographer. Cryptographers have a theoretical advantage over cryptanalysts because the work of a cryptanalysis is exponential in the size of the key.
Once upon a time it was common for cryptographers to rely on secret protocols and crypto algorithms. Cryptanalysts developed an uncanny ability to guess how crypto protocols worked; indeed most good cryptanalysts are ex cryptographers and vice-versa. A significant advantage of the cryptanalyst is that he can conceal his successes, at least if dissemination of the yield of cryptanalysis can be suitably limited. Such limitations also limit the utility of cryptanalysis and so there is a tradeoff. By contrast the cryptographer can not easily ‘test’ his product except by employing the skills of a friendly cryptanalyst which hopefully include the skills of the opposing cryptanalyst. In large institutions devoted to these arts these two groups work together closely.
The art of cryptanalysis spans many disciplines; from telescopes which watch blinking modem lights in distant offices, to monitoring power consumption of a building on a millisecond basis, to introducing false rumors which are likely to be passed over encrypted channels as ‘chosen plaintext’. Cryptography is a narrower art, except as to avoidance of such holes.
Steganography is the art of hiding information within some common redundant data format. Such formats include some noise which steganography can replace with a hidden payload. When advances are made in information compression, steganography must be redone. The steganalyst may be aware of statistical properties of the noise and thereby test files in such formats for hidden payloads. These statistical properties vary depending on the source of the noisy data and it may be difficult to convince a judge that there is encrypted payload. I see no real advantage for either the steganogropher or steganalyst.
As I have followed the technology I have been blindsided several times by developments. When the cryptographer first hears of the chosen plain text attack it seems unreasonable. But chosen plain text has served the cryptanalyst. Chosen cipher text seems even more outrageous either as too difficult to pull-off or too difficult to exploit. Yet crypto systems have fallen to that as well. Authentication seems to be a natural side effect of good crypto, but there is a literature devoted to surprising attacks that get past the obvious authentication techniques. Is there an end to these attack modes? I don’t know. Crypto is hard.