Some papers on some techniques, even modes that make weaker key generation: http://www.cryptovirology.com/cryptovfiles/newbook.html
The last real non NSA approved system was PGP in early 1990's, look up Phil Zimmerman and understand that if you create a system that the NSA has trouble peering into if needed (sometimes with considerable processing power) you will be contacted.
http://www.rsa.com/rsalabs/node.asp?id=2316
"As the premier cryptographic government agency, the NSA has huge financial and computer resources and employs a host of cryptographers. Developments in cryptography achieved at the NSA are not made public; this secrecy has led to many rumors about the NSA's ability to break popular cryptosystems like DES (see Section 3.2), as well as rumors that the NSA has secretly placed weaknesses, called ``trapdoors,'' in government-endorsed cryptosystems. These rumors have never been proved or disproved. Also the criteria used by the NSA in selecting cryptography standards have never been made public."
There would never be definitive evidence and if so it would cause huge problems, for most communications including fast SSL/HTTPS exchanges for business or regular private communications this is not a worry. Most encryption is good enough for that level of communication, there is a reason though the military for sensitive information is using much greater levels and systems for crypto than consumer grade. Just don't be naive to think the NSA can't right now read all of your communications encrypted or not through many avenues trapdoors, backdoors in libraries, taking over computers to get private keys (all keys are stored somewhere and all passwords have to be typed in at some point) etc.