-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 This email is in response to the BugTraq posting at http://online.securityfocus.com/archive/1/299046 There are two issues in the original email which are addressed below. 1) The TCP stack on the PIX is non RFC compliant in responding to TCP packets destined to the network broadcast address. One could craft a telnet/ssh client to connect to the PIX by sending requests to the network broadcast address of the subnet the PIX is connected to. Even if one was able to connect to the PIX, by using such a crafted client, one would still need an account/password to gain access to the PIX. Security of the PIX is not compromised. A router does not allow directed broadcasts by default so such behavior can only be experienced on the local subnet. If directed broadcast is required for a subnet then using the ACL option of the directed broadcast command on the router, TCP directed broadcasts can be filtered out for the subnet. This nonconformant behavior is being fixed in all upcoming PIX releases by allowing new TCP sessions to be created only if the packet was sent to the PIX interface address. Packets sent to the broadcast or subnet address would be dropped. 2) PIX releases unused memory and will allocate memory using a best fit scheme which will reuse freed chunks of memory. When allocating memory, the PIX will first attempt to re-use memory that was freed and not part of the contiguous heap. Cisco has performed additional testing and confirms that no fragmentation or memory leaks are seen based on the attack described in this report. - -- Sharad Ahlawat Cisco Product Security Incident Response Team (PSIRT) http://www.cisco.com/go/psirt Phone:+1 (408) 527-6087 PGP-key: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xC12A996C -----BEGIN PGP SIGNATURE----- Comment: Sharad Ahlawat - PGP Fingerprint: 9A93 2A20 43E5 7F01 2954 C427 1A81 A898 C12A 996C iD8DBQE9z+5uGoGomMEqmWwRAkutAKDrIibzMoFWk/7jNLYxrLnE68Oh8wCgxDWI ZpPJruButb2d+Kz8EIDTHO4= =P3Yk -----END PGP SIGNATURE-----
