2 ms·
Hm, why userspace PRNGs are bad? Does code executed in kernel mode become better automatically? If OpenSSL has bad RNG, then a lot of SSL traffic is in danger.
by avodonosov 10y ago
Hm, why userspace PRNGs are bad? Does code executed in kernel mode become better automatically?
If OpenSSL has bad RNG, then a lot of SSL traffic is in danger...
Also, do you think Java's SecureRandom is not a user space implementation? I guess it only uses /dev/[u]random for seed.
- CiPHPerCoder 10y ago> Hm, why userspace PRNGs are bad? Does code executed in kernel mode become better automatically? At the beginning of the blog post there are three links. adequately -> http://www.2uo.de/myths-about-urandom/ covered -> http://sockpuppet.org/blog/2014/02/25/safely-generate-random-numbers/ elsewhere -> https://blog.cr.yp.to/20140205-entropy.html I highly recommend reading all three. They answer your question in painstaking detail. > Also, do you think Java's SecureRandom is not a user space implementation? I guess it only uses /dev/[u]random for seed. In Java 8, it just reads from urandom. The relevant section has a cautionary note: https://paragonie.com/blog/2016/05/how-generate-secure-random-numbers-in-various-programming-languages#java-csprng https://paragonie.com/blog/2016/05/how-generate-secure-rando...