[radext] Re: Selfie Attack on TLS-PSK

[Fabian Mauchle <[e-mail]>]

On 25.07.2024 03:47, Margaret Cullen wrote:
> Another (D)TLS security factor concerns _when_ RADIUS packets containing 
> privacy-sensitive or security-sensitive data are sent over the (D)TLS 
> session.  We need to make sure that the mechanisms we are using will not 
> transmit the initial RADIUS packet contents until an encrypted session 
> has been established with a properly authenticated and authorized peer.

I think that is part of radiusdtls-bis, and it already says so (section 
3.2 and 4.1).

> I am currently trying to learn enough about DTLS to understand whether 
> that is guaranteed by the DTLS protocol, or if the RADIUS server needs 
> to do something to make sure that happens.
> 
> I would welcome help understanding how this works  from people with good 
> DTLS knowledge and experience…

 From my own implementation, in that regard DTLS is no different than 
TLS. You first open the underlying socket, where for DTLS you just bind 
the UDP socket to the address and port, and then tell your TLS library 
to start a TLS session on that socket.

After that, (for (D)TLS connections you simple never write() or send() 
on the socket itself, but only use the TLS library's write method. The 
TLS library will then handle actually sending and receiving packets and 
make sure the security properties are enforced.

I'm sure there are other ways, but that's how its done in radsecproxy.

-- 
Fabian Mauchle
Network
NOC:    [phone]
Direct: [phone]

Switch
Werdstrasse 2, P.O. Box, 8021 Zurich, Switzerland