Re: [TLS] MITM Attacks on Client Authentication after Resumption

[Andrei Popov <[e-mail]>]

Ø  I would note that the renegotiation handshake we are speaking about is not exactly a standalone authentication protocol. It is the second handshake on a connection, with different client or server identity than the first handshake. This leaves room for confusion: should an implementation associate the connection to the first set of principals or the second?

When a renegotiation happens, the TLS client must validate the newly negotiated parameters, including the server cert, the cipher suites, Finished hash, etc. A TLS client that neglects to validate the server certificate after renegotiation simply has a bug.

It appears that the attack described is only feasible when two implementation defects are present:

1.       TLS implementation error in the client (failure to validate the server cert),  and

2.       A server application that associates data received from an unauthenticated client with the data received after enabling TLS client authentication.

While we cannot prevent the latter at the TLS protocol layer, I believe fixing the bug in the TLS client effectively thwarts this attack.

Thanks,

Andrei

From: TLS [mailto:[e-mail]] On Behalf Of Karthikeyan Bhargavan
Sent: Monday, March 3, 2014 4:47 PM
To: Paterson, Kenny
Cc: [e-mail]
Subject: Re: [TLS] MITM Attacks on Client Authentication after Resumption


Rather, an application running at C might end up using TLS-protected data
that was supplied by M in step 2 of the attack as if it came from S (or
vice-versa). This is a comparable attack to the original renegotiation
attack of Ray&Dispensa and Rex.

Yes, indeed, the impact of our renegotiation attack is comparable to the 2009 renegotiation attacks.
In fact, the setup of our attack is closer to Rex's version than Ray's version:
http://www.ietf.org/mail-archive/web/tls/current/msg03928.html

as far as I can see - for the 3rd step of the attack, it
seems that M just passes messages backwards and forwards and ends up NOT
knowing the keys shared by C and S. Any authentication protocol is
vulnerable to a "message passing" attack of this type in which the
adversary simply acts as a wire, and it's not considered an attack on such
protocols.

I would note that the renegotiation handshake we are speaking about is not exactly a standalone authentication protocol. It is the second handshake on a connection, with different client or server identity than the first handshake. This leaves room for confusion: should an implementation associate the connection to the first set of principals or the second?

Indeed, it's correct that C *is* connected to S and S *is* connected to C
after step 3! (not just that they "think" they are so-connected).

Kenny, you seem to suggest that only the second should be considered. Common web browsers and HTTPS libraries only consider the first. It is interesting to try and understand which is the right answer here.

-Karthik




What matters here is how a server application running at S interprets data
that may have been sent under TLS protection in the first 2 steps in the
attack: it may assume it came from C when in fact it came from M.

The more accurate description from Karthik's text is this:


Step 3. (Renegotiation C-M-S)


...
Both handshakes complete with new mutually-authenticated sessions and
record keys.
C now thinks it is connected to S and S thinks it is connected to C.
(M does not know the new record keys but its previous messages to S on
the same connection
may be treated as authenticated by C.)


Indeed, it's correct that C *is* connected to S and S *is* connected to C
after step 3! (not just that they "think" they are so-connected).

Cheers

Kenny


On 03/03/2014 15:20, "Karthikeyan Bhargavan"
<[e-mail]<mailto:[e-mail]>> wrote:

We are going to present some new man-in-the-middle attacks on TLS
applications in Tuesday¹s working group meeting. These attacks were
found as part of our research on trying to prove cryptographic
security for a TLS implementation [1].  We presenting some of the
materials here to kick-start some discussion and get early comments on
our proposed countermeasure. Some people on the list already know of
these attacks but this is the first public disclosure.

More details are at https://secure-resumption.com
<https://secure-resumption.com/> [2]

Scenario
======
Consider a client C that normally authenticates to a server S using a
client certificate.  If C uses the same certificate to authenticate to
a malicious server M, then we show that M can use C¹s certificate to
authenticate its own connection to S.

The attack relies on the combination of an initial RSA or DHE
handshake, followed by session resumption on a new connection,
followed by a client-authenticated renegotiation. During the first two
handshakes, C has a connection to M and M has a connection to
S. During the third handshake, M is able to authenticate as C to S and
as S to C.

This server-based man-in-the-middle attack should normally have been
prevented by the Renegotiation Indication (RI) extension [3] but by
injecting session resumption between the two full handshakes, we are
able to bypass the renegotiation countermeasure.

Triple Handshake Attack
================
I¹ll briefly summarise the attack below for an initial RSA key
exchange.  The webpage [2] has diagrams that will be easier to follow,
describes more attack variants, and provides some disclosure status.

The attack proceeds in three steps:

Step 1. (Initial Handshakes C-M, M-S)
- C connects to M and M connects to S, both handshakes use RSA.
- M forwards C¹s and S¹s client hellos to each other.
- M receives an encrypted PMS from C and reencrypts it towards S.
- Both handshakes complete with new sessions and record keys.
 Both sessions have the same master secret, random nonces, and session id.
 (M knows the master secret and record keys since it participated in both
handshakes)

Step 2. (Session Resumption C-M, M-S)
- C resumes its session with M on a new connection.
- M resumes its session with S on a new connection.
- M forwards all the abbreviated handshake messages unchanged between C
and S.
- Note that the RI extensions on both handshakes are empty,
 since it is the first handshake on the connection
- Both handshakes complete with new record keys (and reuse old sessions)
 Both connections have the same record keys and handshake logs (verify
data)
 (M still knows the record keys and can send messages in either
direction.)

Step 3. (Renegotiation C-M-S)
- S requests M for renegotiation with client certificate.
 M requests C for renegotiation with client certificate.
- M forwards all renegotiation messages unchanged between C and S
- Note that since the handshake logs in the preceding handshake were the
same,
 the RI extensions on both handshakes will be the same.
- Both handshakes complete with new mutually-authenticated sessions and
record keys.
 C now thinks it is connected to S and S thinks it is connected to C.
 (M does not know the new record keys but its previous messages to S on
the same connection
 may be treated as authenticated by C.)

At the end of Step 3, S has an incoming connection on which it
initially received data from an anonymous client (M) and later
received data from an authenticated client (C). This breaks the
intended guarantees of the RI extension.

Countermeasures
===========
During Step 3, C has a connection on which it first received M¹s
certificate and later S¹s certificate. If C refuses to accept this
change of server identity, then it can prevent Step 3 of the
attack. Indeed, we recommend mainstream web browsers and HTTPS
libraries should systematically forbid the change of server identities
during renegotiation.

However, already at the end of Step 2, a number of connection and
session parameters, such as the tls-unique channel binding for the two
connections are the same. So any application-level mechanism that
relies on the TLS master secret [4] or channel bindings [5] or exports
TLS keying material [6] is vulnerable to a similar man-in-the-middle
attack.

We argue that the core vulnerability here is that the TLS master
secret is not bound to enough elements of the TLS session. We propose
a new TLS extension that binds the master secret to the hash of the
all relevant handshake messages in the initial handshake.

The proposed draft is available at:
http://secure-resumption.com/draft-bhargavan-tls-session-hash-00.txt

The key idea is that each full handshake is associated with a session
hash, computed as

session_hash = Hash(handshake_messages)

where handshake_messages consist of all messages up to and including
the ClientKeyExchange.  The extended master secret computation enabled
by the extension is then computed as

master_secret = PRF(pre_master_secret,
                                       "extended master secret",
                                        session_hash) [0..47];

We¹ve implemented this extension in OpenSSL without much difficulty.
Changing the master secret derivation may seem radical, but we believe
it is the main way to counter future attacks that may rely on the
session synchronization (step 1) that we exploit here.

An alternative countermeasure would be an extension (along the lines
of [3]) that includes the session hash as defined above in the
ClientHello and ServerHello messages of the abbreviated
handshake. This would provide an explicit link between the resumption
handshake and its original full handshake, and hence prevent the
renegotiation attack described above.

We welcome comments and suggestions.
-Karthik Bhargavan, Antoine Delignat-Lavaud, and Alfredo Pironti

[1] http://mitls.org <http://mitls.org/>
[2] https://secure-resumption.com <https://secure-resumption.com/>
[3] RFC5746: Transport Layer Security Renegotiation Indication Extension
[4] The Compound Authentication Binding Problem
(draft-puthenkulam-eap-binding-04)
[5] RFC5929: Channel Bindings for TLS
[6] RFC5705: Keying Material Exporters for Transport Layer Security