Installation guide

Bugzilla 351162
RHBA-2009:1596 This update addresses Bug
505682 - Allow configuration of
NSS OCSP cache settings. New
parameters are enabled to allow
user-defined cache sizes,
OCSP check times, and timeout
periods for OCSP responses.
November 19, 2009
RHBA-2009:1443 This release had enhancements
for ECC support, including
extending support on Firefox for
ECC enrollments and adding
support for ECC POP. T his
release also included these bug
fixes:
Bugzilla 512831
Bugzilla 512828
Bugzilla 513450
Bugzilla 514093
Bugzilla 514270
Bugzilla 518431
September 14, 2009
8. Known Issues
8.1. Reconfiguring the Red Hat Certificate System Subsystems to Prevent a
Potential TLS-Related Man-in-the-Middle Attack
Transport Layer Security (TLS) is a protocol which establishes a secure connection between a client
and a server. Marsh Ray of PhoneFactor discovered a flaw in the TLS protocol itself which could allow
an attack to insert plain text into an existing session during a TLS renegotiation operation.
The Educated Guesswork blog has a good description of this kind of attack at
http://www.educatedguesswork.org/2009/11/understanding_the_tls_renegoti.html.
Either a client or a server may request a renegotiation of an existing TLS/SSL session (for instance, to
renew session encryption keys or to use different cipher suite). When T LS/SSL is used to secure
access to an HTTP service and a client attempts to access some protected resource, server-initiated
renegotiation asks client to authenticate with a certificate.
However, the TLS/SSL protocols did not use any mechanism to verify that session peers do not change
during the session renegotiation. T herefore, a man-in-the-middle attacker could use this flaw to open
TLS/SSL connections to the server, send attacker-chosen request to the server, trigger the renegotiation
(either by directly requesting it or by attempting to access protected resource, resulting in server-
initiated renegotiation) and splice victim's initial connection attempt to an existing TLS/SSL session.
Depending on the application-layer protocol, this may lead to attacker request being performed by the
server as if authenticated using victim's credentials or using data from victim's request. After the
renegotiation, attacker can no longer decrypt communication between the client and the victim, so this
attack is also referred to as a "blind prefix injection" attack. Eric Rescorla's blog post "Understanding the
Red Hat Certificate System 8.0 Red Hat Certificate System 8.0
22