6.2p1+ fails key-exchange with embedded SSH devices

[TJ]

I have various network devices with embedded SSH daemons, such as CDUs, UPSs, managed Switches, DRAC, and so on. All devices have entries in the ssh config file specifying an IdentityFile.

The particular embedded SSH daemon in this case is Mocana nanoSSH:

http://www.mocana.com/for-device-manufacturers/nanossh

Recently I had need to access a Server Technology Switched CDU via SSH using 14.04 Trusty and its OpenSSH_6.6.1p1. Connections consistently failed at:

$ ssh -vvv cdu-01
OpenSSH_6.6.1, OpenSSL 1.0.1f 6 Jan 2014
debug1: Reading configuration data /home/tj/.ssh/config
debug1: /home/tj/.ssh/config line 101: Applying options for cdu-01
debug1: Reading configuration data /etc/ssh/ssh_config
debug1: /etc/ssh/ssh_config line 19: Applying options for *
debug1: Hostname has changed; re-reading configuration
debug1: Reading configuration data /home/tj/.ssh/config
debug1: Reading configuration data /etc/ssh/ssh_config
debug1: /etc/ssh/ssh_config line 19: Applying options for *
debug2: ssh_connect: needpriv 0
debug1: Connecting to 10.254.0.3 [10.254.0.3] port 22.
debug1: Connection established.
...
debug2: kex_parse_kexinit: reserved 0
debug2: mac_setup: setup hmac-md5
debug1: kex: server->client aes128-cbc hmac-md5 none
debug2: mac_setup: setup hmac-md5
debug1: kex: client->server aes128-cbc hmac-md5 none
debug2: bits set: 531/1024
debug1: sending SSH2_MSG_KEXDH_INIT
debug1: expecting SSH2_MSG_KEXDH_REPLY
Received disconnect from 10.254.0.3: 11: Logged out.

Initially I thought the local configuration had been corrupted in some way since the local /home/ was moved from a 13.04 to 14.04 installation. Later I discovered a bug report of this issue affecting other embedded network devices reported against versions of OpenSSH later than 6.1p1 on the BSD operating system at:

https://bugzilla.mindrot.org/show_bug.cgi?id=2116

I then chroot-ed into the earlier 13.04 installation and tried again with its OpenSSH_6.1p1 and it connected:

$ ssh -vvv hostmaster@10.254.0.3
OpenSSH_6.1p1 Debian-4, OpenSSL 1.0.1c 10 May 2012
debug1: Reading configuration data /etc/ssh/ssh_config
debug1: /etc/ssh/ssh_config line 19: Applying options for *
debug2: ssh_connect: needpriv 0
debug1: Connecting to 10.254.0.3 [10.254.0.3] port 22.
debug1: Connection established.
...
debug2: kex_parse_kexinit: reserved 0
debug2: mac_setup: found hmac-md5
debug1: kex: server->client aes128-cbc hmac-md5 none
debug2: mac_setup: found hmac-md5
debug1: kex: client->server aes128-cbc hmac-md5 none
debug2: dh_gen_key: priv key bits set: 117/256
debug2: bits set: 521/1024
debug1: sending SSH2_MSG_KEXDH_INIT
debug1: expecting SSH2_MSG_KEXDH_REPLY
debug1: Server host key: DSA 17:36:55:87:49:a5:a4:9f:02:42:8b:da:88:7c:bb:41
The authenticity of host '10.254.0.3 (10.254.0.3)' can't be established.
DSA key fingerprint is 17:36:55:87:49:a5:a4:9f:02:42:8b:da:88:7c:bb:41.
Are you sure you want to continue connecting (yes/no)?

The cause is apparently a buffer overflow in the embedded Mocana SSH daemon with cipher/mac lists longer than ~287 bytes.

I fixed it by limiting the list of Ciphers and Macs sent, via the ssh_config:

Host cdu-01
Hostname 10.254.0.3
User hostmaster
IdentityFile ~/.ssh/id_hostmaster
IdentitiesOnly yes
Protocol 2
Ciphers aes256-cbc,aes192-cbc,aes128-cbc,blowfish-cbc,3des-cbc,arcfour
Macs hmac-sha1,hmac-md5

$ ssh -vvv cdu-01
OpenSSH_6.6.1, OpenSSL 1.0.1f 6 Jan 2014
debug1: Reading configuration data /home/tj/.ssh/config
debug1: /home/tj/.ssh/config line 101: Applying options for cdu-01
debug3: ciphers ok: [aes256-cbc,aes192-cbc,aes128-cbc,blowfish-cbc,3des-cbc,arcfour]
debug3: macs ok: [hmac-sha1,hmac-md5]
debug1: Reading configuration data /etc/ssh/ssh_config
debug1: /etc/ssh/ssh_config line 19: Applying options for *
debug1: Hostname has changed; re-reading configuration
debug1: Reading configuration data /home/tj/.ssh/config
debug3: ciphers ok: [aes256-cbc,aes192-cbc,aes128-cbc,blowfish-cbc,3des-cbc,arcfour]
debug3: macs ok: [hmac-sha1,hmac-md5]
debug1: Reading configuration data /etc/ssh/ssh_config
debug1: /etc/ssh/ssh_config line 19: Applying options for *
debug2: ssh_connect: needpriv 0
debug1: Connecting to 10.254.0.3 [10.254.0.3] port 22.
debug1: Connection established.
...
debug1: sending SSH2_MSG_KEXDH_INIT
debug1: expecting SSH2_MSG_KEXDH_REPLY
debug1: Server host key: DSA 17:36:55:87:49:a5:a4:9f:02:42:8b:da:88:7c:bb:41
debug3: load_hostkeys: loading entries for host "10.254.0.3" from file "/home/tj/.ssh/known_hosts"
debug3: load_hostkeys: loaded 0 keys
The authenticity of host '10.254.0.3 (10.254.0.3)' can't be established.
DSA key fingerprint is 17:36:55:87:49:a5:a4:9f:02:42:8b:da:88:7c:bb:41.
Are you sure you want to continue connecting (yes/no)?

[Seth Arnold (seth-arnold)]

It might be worth changing your ciphers to something like:

Ciphers AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA

and see if you can get 0x41414141 kinds of things in e.g. CPU registers
or elsewhere in your embedded devices. (I'm curious if this is just a
client configuration that won't connect properly or if you've found
exploitable flaws in the Mocana sshd.)

Thanks

[TJ (tj)]

On 31/07/14 04:18, Seth Arnold wrote:
> It might be worth changing your ciphers to something like

Seth, that is terrible .... OK, I'll try it later on, although I have no idea what CPU the embedded controller on the CDU uses. If I knew that I might be able to craft the buffer overflow to do
something useful.

I've been explicit in describing the issue since it looks like the kind of thing that many sys-admins could hit and not get to the bottom of.

[Durai (b-duraai)]

'ssh -o HostKeyAlgorithms=ssh-rsa,ssh-dss -o KexAlgorithms=diffie-hellman-group1-sha1 -o Ciphers=aes128-cbc,3des-cbc -o MACs=hmac-md5,hmac-sha1 user@host' only works. please let me know what is the issue here.