4 ms·
The linked PDF: http://www.chronox.de/lrng/doc/lrng.pdf http://www.chronox.de/lrng/doc/lrng.pdf Page 4: "The new approach implements the modeling of a slow noi
by mikepurvis 11y ago
The linked PDF: http://www.chronox.de/lrng/doc/lrng.pdf http://www.chronox.de/lrng/doc/lrng.pdf
Page 4: "The new approach implements the modeling of a slow noise source within the
LRNG based on the timing of interrupts and allowing other, even fast operating
noise sources to contribute entropy. As discussed above for the legacy /dev/random,
only the high-resolution time stamps deliver entropy for hardware events
and other information processed by the legacy /dev/random implementation
hardly have any entropy. This identification is used as a basis that solely the
timing of interrupts is used."
- acqq 11y ago"The cryptographic processing of the entropic data is implemented with a well-known and well-studied deterministic random number generator: an SP800-90A DRBG as defined in [2] – naturally excluding the NSA-sponsored Dual-EC DRBG." "The concept of the LRNG allows the selection of the used DRBG and the backend cipher implementations at compile time." Is SP800-90A DRBG really state of the art? According to Wikipedia "the publication contains the specification for four cryptographically secure pseudorandom number generators for use in cryptography." Without Dual-EC it's still three different DRBGs. How good are they? Is there anything better?
- tptacek 11y agoThey're fine. They're sane, straightforward, relatively simple constructions based on hashes, HMAC, and ciphers running in CTR mode. You could make a CSPRNG more complicated, but there's not much reason to. The stuff that extends the "state of the art" beyond SP800-90A tends to be system-specific and feature-based. For instance, the "estimators" in the LRNG are an extension to SP800-90A (and a bad idea). Rolling over seed files are another extension.