lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [day] [month] [year] [list]
Date: Sat, 6 Jun 2015 17:22:13 +0530
From: Payatu Research <research@...atu.com>
To: bugtraq@...urityfocus.com
Subject: Hardcoded AES 256 bit key used in Kankun IoT/Smart socket and its
 mobile App

Hi List,

Vulnerability
=============
Hardcoded AES 256 bit key used in Kankun IoT/Smart socket and its mobile App

Vulnerability Description
==========================
The kankun smart socket device and the mobile app use a hardcoded AES
256 bit key to encrypt the commands and responses between the device
and the app. The communication happens over UDP. An attacker on the
local network can use the same key to encrypt and send unsolicited
commands to the device and hijack it.

CVE ID
========
CVE-2015-4080

Vendor
========
www.ikonke.com

Product
=========
Kankun Smart Socket

Disclosure Timeline
====================
1. 25 May 2015 – Reported to Vendor, no response.
2. 29 May 2015 – Reminder sent to vendor, no response.
3. 5 June 2015 – Public disclosure.

Credits
=========
1. Aseem Jakhar, Director - Research, Payatu Technologies Pvt. Ltd.
2. Since at the time of publishing the finding, we searched online for
the same and found that someone else had also published the key. In
good faith we would like to mention the same person who goes by the
handle: kankun hacker -
https://plus.google.com/109112844319840106704/posts although both the
research were independent of each other and we do not know who kankun
hacker is.

About Payatu
============
Payatu Technologies is a boutique security testing company. We
specialize in Mobile/IoT/Product/Application security testing.

PoC exploit source code
========================
https://bitbucket.org/aseemjakhar/kcmd

Technical details
==================

We performed our analysis on the Android App and the device. The user
manual specifies the app to be used for the device -
http://kk.huafeng.com:8081/none/android/smartwifi.apk The smart socket
has a newer version on the app on google play store which is also
vulnerable - https://play.google.com/store/apps/details?id=hangzhou.zx


Communication
--------------
The communication between the app and the device happens over UDP.
The commands are Broadcasted on the network to UDP destination port 27431


App Reversing
----------------
- We decompiled the app using using apktool
- The app has a native shared library libNDK_03.so which contains the
encryption logic and the hard-coded key
- We analysed the app and got an idea of the command/response protocol
being used between the app and the device.
- The Java code uses JNI functions to encrypt and decrypt the commands
and responses. The functions are encode() and decode(). Interestingly
there is also a function called add() which adds the two parameters
and returns the result. This must definitely be a testing function
used while starting to develop the library :).

- The command and response for switching ON and OFF is a 4 step process
  - Step 1 - App sends an Open/Close request (Open means Switch ON,
close means Switch OFF)
  - Step 2 - Device sends a response containing a confirmation ID (a number)
  - Step 3 - App sends the confirmation request along with the
confirmation ID received from the Device
  - Step 4 - Device sends an Acknowledgement and Switches the device ON/OFF
  - An example of the communication protocol to Switch ON the device,
assuming the MAC address of the device is “de:ad:de:ad:de:ad”, the
password set by the user is “secretpass” and the confirmation ID is
70018. If the user does not set any encryption password the string
“nopassword” is used.

  APP --> lan_phone%de:ad:de:ad:de:ad%secretpass%open%request --> Device
  DEVICE -->  lan_device%de:ad:de:ad:de:ad%secretpass%confirm#70018%rack --> APP

  APP -->  lan_phone%de:ad:de:ad:de:ad%secretpass%confirm#70018%request
--> DEVICE
  DEVICE --> lan_device%de:ad:de:ad:de:ad%secretpass%open%rack --> APP

  - As you can see above, the communication is a simple string where
fields are separated by the % character. The fields are self
explanatory. There is also an option of wan_phone and wan_device which
we did not test.

- A quick strings output showed up the key along with other strings.
This particular string looked a little intersting and we started
reversing the native library.
- Output of $ strings libNDK_03.so

....
UUPx((
Zw--
fdsl;mewrjope456fds4fbvfnjwaugfo
java/lang/String
....


- The installed library is not stripped
- In the library the JNI encode/decode functions call
EncryptData/DecryptData respectively

00003990 <Java_hangzhou_kankun_WifiJniC_encode>:
    3990:       b5f0            push    {r4, r5, r6, r7, lr}
    3992:       b085            sub     sp, #20
    3994:       1c11            adds    r1, r2, #0
    ....
    39b6:       2380            movs    r3, #128        ; 0x80
    39b8:       f7ff ff0e       bl      37d8 <EncryptData>
     ....

- The EncryptData() internally calls aes functions which means it is
using AES encryption

000037d8 <EncryptData>:
    37d8:       b5f0            push    {r4, r5, r6, r7, lr}
    37da:       465f            mov     r7, fp
    37dc:       4656            mov     r6, sl
    ....
    3868:       9001            str     r0, [sp, #4]
    386a:       f7fd fb8b       bl      f84 <aes_set_key>
    ....
    38be:       1c29            adds    r1, r5, #0
    38c0:       f7fd fd84       bl      13cc <aes_encrypt>

- Now to find the key we look at the code of EncryptData(). The below
code generates the memory address of the key string within the
library.

    3842:       4b4e            ldr     r3, [pc, #312]  ; (397c
<EncryptData+0x1a4>)
    ....
    384a:       447b            add     r3, pc
    ....
    384e:       3328            adds    r3, #40 ; 0x28
    ....
    397c:       00003e96        muleq   r0, r6, lr

- Lets look at the code:
  - The ldr instruction loads the value 0x3e96 in r3, which is the
value at address 0x397c.
  - The “add r3, pc” instruction adds 384e to r3. Because of
pipelining and THUMB mode while executing the “add r3, pc” instruction
pc will point 4 bytes ahead of the address 0x384a i.e. 0x384e.
  - Finally the instruction “adds r3, #40” adds 0x28 (decimal 40) to r3.
  - In effect r3 = 0x3e96 + 0x384e + 0x28 = 0x770c which is where our
key resides.
  - The key is fdsl;mewrjope456fds4fbvfnjwaugfo

    770c:       6c736466    # fdsl
    7710:       77656d3b   # ;mew
    7714:       706f6a72    # rjop
    7718:       36353465   # e456
    771c:       34736466    # fds4
    7720:       66766266    # fbvf
    7724:       61776a6e     # njwa
    7728:       6f666775     # ugfo

- The algorithm uses 256 bits which is set in register r2 before the
call to aes_set_key(). The below instructions translate to r2 = 128
Left shift by 1 = 256
    3860:       2280            movs    r2, #128        ; 0x80
    ....
    3866:       0052            lsls    r2, r2, #1

- The same function i.e. aes_set_key() is called from DecryptData() as well.
- We also confirmed that the key is being used by the UDP service
listening on the device by simply running strings on it and we could
see the same string in the program. There are two binaries in the
device that use the key - kkeps_off and kkeps_on
- Output of $ strings kkeps_off

....
B$Bh0
pointer is null
pucOutputData too small
fdsl;mewrjope456fds4fbvfnjwaugfo
pucInputData dataLen is incorrect
pucOutPutData is too small
....

Hijacking the Smart Socket
---------------------------
Now that we have the encryption key lets go hijack the socket. Instead
of writing your own openssl wrapper for encryption/decryption, we can
simply use the native shared library to do the magic for us. A quick
look at the IDA output shows that EncryptData() and DecryptData() are
exported functions which means any native program can load libNDK_03
using dlopen() and family and can use these functions which is exactly
what we did.

IDA output
.text:000037D8 ; int EncryptData(unsigned __int8 *pucInputData, int
nInPutLen, unsigned __int8 *pucOutputData, int nOutputsize, int
*pnOutPutLen)
.text:000037D8                 EXPORT EncryptData

But wait a minute! What if the user sets a password :O. Since, there
is an option called Encryption in the app (which actually means set a
password to be used between the app and the device communication and
not actually an encryption passphrase), if the user sets a password,
only that user will be able to control the device. Worry not! We
previously saw that the app sends the commands as UDP broadcasts to
port 27431. Which means anyone on the local network can sniff the
command packets. So all you need to do is:
1. nmap scan for UDP port 27431 on the network. When you find the
device note down its MAC address to be used in the request.
2. Encrypt and Send open/close request with default password string
i.e “nopassword”
3. Check the response
1. If response received with confirmation ID goto step 7
2. If no response received or response unknown goto step 4
4. Listen on (broadcast address) UDP port 27431 and sniff all requests.
5. Decrypt the request received and extract the password.
6. Encrypt and send open/close request with the extracted password.
7. Extract the confirmation ID from the response
8. Send the confirmation request to the device
9. Receive the acknowledgement response
10. The device has been successfully hijacked

Regards,
Payatu Research Team

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ