Hi, https://www.ssllabs.com/ssltest/analyze.html?d=ruggedinbox.com report is not good enough, your should specify more strict cipher policy, to have full Forward Secrecy. Roundcude is insecure (many private exploits available), but I like it better then squirrel.
And finally after registration, I can't login. (maybe you not support some special symbols in passwords, used 21 long). With normal Forward Secrecy self-signed certificates is just piece of useless crap. Don't forget not only select long dh param/key, but to change default curve for at least longer one(don't know is it possible or not with lighttpd).
And finally after registration, I can't login. (maybe you not support some special symbols in passwords, used 21 long). With normal Forward Secrecy self-signed certificates is just piece of useless crap. Don't forget not only select long dh param/key, but to change default curve for at least longer one(don't know is it possible or not with lighttpd).
Hi cryptofutureis, thanks for your detailed suggestions about ssl!
By following this howto (forward secrecy on lighttpd): https://raymii.org/s/tutorials/Strong_SSL_Security_On_lighttpd.html
score raises to A
with this parameters: https://cipherli.st
the overall rating is A+
https://www.ssllabs.com/ssltest/analyze.html?d=ruggedinbox.com
easy and very useful!
(also, today those debian packages: libssl-dev libssl-doc libssl1.0.0 libssl1.0.0:i386 openssl were updated)
About the password, we made some (manual) tests and the invalid characters are " (quote) and \ (back-slash aka 'reverse solidus')
so you can have passwords like `~!@#$%^&*()-=_+}{[];'
and ,./<>?
we didn't test symbols, anyway the only character that we really strip is " (quote)
About Roundcube, now that you say that (0-day exploits available around), you gave us the additional motivation to configure spawn-fcgi to isolate the virtual hosts (so hacking roundcube would not result in having access to the whole document root of the web server) .. we'll do that as the next thing.
Thanks for your feedback and happy emailing!
Diffie-Hellman and Elliptic-Curve Diffie-Hellman key agreement protocols will be supported in lighttpd 1.4.29. By default, Diffie-Hellman and Elliptic-Curve Diffie-Hellman key agreement protocols use, respectively, the 1024-bit MODP Group with 160-bit prime order subgroup from RFC 5114 and "prime256v1" (also known as "secp256r1") elliptic curve from RFC 4492. The Elliptic-Curve Diffie-Hellman key agreement protocol is supported in OpenSSL from 0.9.8f version onwards. For maximum interoperability, OpenSSL only supports the "named curves" from RFC 4492.
Using the ssl.dh-file and ssl.ec-curve configuration variables, you can define your own set of Diffie-Hellman domain parameters. For example:
ssl.dh-file = "/etc/lighttpd/ssl/dh2048.pem"
ssl.ec-curve = "secp384r1"
Default is secp256r1 but we always can select curve with bigger prime.
Mozilla has a nice doc available: https://wiki.mozilla.org/Security/Server_Side_TLS