Shib IDP's LDAPS attribute resolution and SSLv3

[Ian Young]

On 16 Oct 2014, at 07:23, Rhys Smith <Smith at cardiff.ac.uk> wrote:

> connections, from what I understand the POODLE attack requires:
> a) the attacker being able to intercept traffic
> b) a client that will downgrade from TLS1 to SSLv3

The padding oracle attack should be possible whenever SSLv3 is negotiated. In the modern world this would normally not happen unless a downgrade dance is forced by the attacker, but if a client only supports SSLv3 then SSLv3 will be negotiated without a downgrade, and as far as I can tell the padding oracle attack is still applicable in that case.

> c) to be able to inject data into the connection to make the client retry.

Bottom line: SSLv3 is no longer secure under your conditions A and C; B is just icing on the cake.

As the only possible reason to continue to support SSLv3 on the server side is to support clients which don't support anything newer than SSLv3, you can therefore ignore condition B entirely.

> Not saying people shouldn’t be considering disabling SSLv3 everywhere now, just that I don’t think there are really known attack vectors for poodle in the ldaps circumstance - yet.

If you have a secure network between the server and all clients, that's true. However, in those circumstances you should feel equally safe without using SSL/TLS at all, no? If you don't feel that safe, you really need to disable SSLv3.

	-- Ian



-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 4877 bytes
Desc: not available
Url : http://shibboleth.net/pipermail/users/attachments/20141016/c75d9b96/attachment.bin