Kerberos is a networked authentication system which uses a trusted third party (a KDC) to authenticate clients and servers to each other.
A double-free flaw was found in the krb5_recvauth() routine which may be triggered by a remote unauthenticated attacker. Although no exploit is currently known to exist, this issue could potentially be exploited to allow arbitrary code execution on a Key Distribution Center (KDC). The Common Vulnerabilities and Exposures project assigned the name CAN-2005-1689 to this issue.
Daniel Wachdorf discovered a single byte heap overflow in the krb5_unparse_name() function, part of krb5-libs. Sucessful exploitation of this flaw would lead to a denial of service (crash). To trigger this flaw an attacker would need to have control of a kerberos realm that shares a cross-realm key with the target, making exploitation of this flaw unlikely. (CAN-2005-1175).
Gaël Delalleau discovered an information disclosure issue in the way some telnet clients handle messages from a server. An attacker could construct a malicious telnet server that collects information from the environment of any victim who connects to it using the Kerberos-aware telnet client (CAN-2005-0488).
The rcp protocol allows a server to instruct a client to write to arbitrary files outside of the current directory. This could potentially cause a security issue if a user uses the Kerberos-aware rcp to copy files from a malicious server (CAN-2004-0175).
All users of krb5 should update to these erratum packages which contain backported patches to correct these issues. Red Hat would like to thank the MIT Kerberos Development Team for their responsible disclosure of these issues.