Inside a low budget consumer hardware espionage implant

The following analysis was performed on a S8 data line locator which replied to the hidden SMS command for version query (*3646655*) with:

Ver=MTK6261M.T16.17.01.10

build=2017/01/10 17:33

Introduction

A while back Joe Fitz tweeted about the S8 data line locator1. He referred to it as “Trickle down espionage” due to its reminiscence of NSA spying equipment.

The S8 data line locator is a GSM listening and location device hidden inside the plug of a standard USB data/charging cable. It supports the 850, 900, 1800 and 1900 MHz GSM frequencies.

Its core idea is very similar to the COTTONMOUTH product line by the NSA/CSS [1] in which an RF device is hidden inside a USB plug. Those hidden devices are referred to as implants.

The device itself is marketed as a location tracker usable in cars, where a thief would not be able to identify the USB cable as a location tracking device. Its malicious use-cases can, however, not be denied. Especially since it features no GPS making its location reporting very coarse (1.57 km deviation in my tests). It can, e.g., be called to listen to a live audio feed from a small microphone within the device, as well as programmed to call back if the sound level surpasses a 45 dB threshold. The fact that the device can be repackaged in its sliding case, after configuring it, i.e. inserting a SIM, without any noticeable marks to the packaging suggests its use-case: covert espionage.

S8 data line locator capabilities

The S8 data line locator has several eavesdropping, espionage and spying capabilities. A SMS message log could look like this:

Listen in

Calling the S8 data line locator for 10 seconds establishes a call and allows you to listen to the microphone feed from the device.

Call back

Sending 1111 via SMS to the device enables voice activated call back. It is acknowledged via the following SMS reply:

DT: Set voice monitoring, voice callback and sound sensitivity:400

Once the audio level goes above 40 dB the device calls back the number that send the 1111 command.

Sending 0000 disables the audio triggered call back. It is replied by:

DT: Voice monitoring cancelled successfully.

Query location

Accoding to the manual sending ‘dw’ via SMS to the device yields a reply SMS with the location. This reply is in the form:

Loc:Street, ZIP City, Country

http://gpsui.net/u/xxxx Battery: 100%

The ‘xxxx’ are replaced with characters ‘0-9,A-Z,a-z’ and the Street, ZIP City, Country line with the appropriate street, ZIP, city and country. The link to http://gpsui.net can be accessed without authorization. It forwards to Google maps.

The location was never more accurate than 1.57 km off.

During the query the device will use a mobile data connection to an unknown endpoint (presumably gpsui.net). This is confirmed by a “MMS/Internet” charge by my provider. My provider does not discern MMS and Internet, but it is save to assume there is an Internet connection established during location query.

This issue was the stepping stone for this analysis. Because the device sends unknown data to an unknown third party it can not - at least with a clear conscious - be used in, e.g., a penetration test. You simply can not use a potentially pre-owned tool.

I therefore tried to analyze and eliminate this phone-home “feature”.

Hardware

To gain access to the devices innards we first tear of the metal shield of the USB connector: S8 data line locator removing USB connector shield.

Next, we remove the plastic cover: S8 data line locator removing USB connector shield.

Chips

After opening the device we can identify the chips:

It features:

• MediaTek MT6261MA: Low budget chip often used in cheap Chinese smartwatches. No official documentation nor information about the chip is available from MediaTek.
• RDA 6626e: “a high-power, high-efficiency quad-band front-end Module […] designed for GSM850, EGSM900, DCS1800, PCS1900 handheld digital cellular equipment.”

Connections

So far I could identify 3 different avenues to connect to the device:

USB (passthrough)

The USB A-connector and the Micro-B cable are not connected to the MT6261MA. They merely pass the signal from one to the other:

UART

The next connection is a UART:

Interfacing with it yields, approximately 3 seconds after booting the device:

screen /dev/ttyUSB0  # 8N1

F1:
V0:   []
:
:
U0:   []
G0:   []
T0:  0C73
Jump to BL

~~~ Welcome to MTK Bootloader V005 (since ) ~~~
**===================================================**

Bye bye bootloader, jump to=0x1000a5b0

However, the output stops there. Input to the device is ignored.

It is likely there exist a different firmware version that accepts AT modem commands2. The boot banner of that alternate firmware references “ZhiPu”3 (some file names of the FAT12 in the firmware flash of my device contain this string as well, so the device firmware is likely related to that other firmware).

USB (MTK)

The DP and DM pads on the USB connector are not connected to the D+ and D- lines of the USB connector. However, the V and GND pads are. The DP and DM pads are instead routed to the MT6261MA processor as illustrated here:

Next, a USB cable must be soldered to these connectors as follows:

The device will then be recognized as an MediaTek phone USB endpoint with the following data:

ID 0e8d: MediaTek Inc. MT6227 phone

This is often called the “MTK boot repair”, “MTK DM DP flash”, etc. It will allow us to interface with the device and dump the firmware ROM and flash.

Dumping firmware

To dump the firmware I use the open source Fernvale research OS [2]. It was initially targeted for the MT6260 processor. It has, however, been ported to the MT6261 and also works on the MT6261MA.

Obtaining and building fernly’s MT6261 branch

A suitable fork of fernly by Urja “urjaman” Rannikko can be obtained and build as follows:

git clone https://github.com/urjaman/fernly
git clone https://github.com/robertfoss/setup_codesourcery.git
sudo setup_codesourcery/setup.sh
/usr/local/bin/codesourcery-arm-2014.05.sh
cd fernly
git checkout fernly6261
make CROSS_COMPILE=arm-none-eabi-
exit
cp -fernvale-simple.rules /etc/udev/rules.d/.

Dumping ROM

To dump the flash we run:

echo "data = [" > rom.py
fernly/build/fernly-usb-loader /dev/fernvale fernly/build/dump-rom-usb.bin >> rom.py
echo "
]
f = open('rom.bin','wb')
for s in data:
  f.write(chr(int(s,)))
f.close()
" >> rom.py
python rom.py

The file rom.bin will now (at least according to the fernly repository documentation) contain the devices ROM.

Dumping flash

To dump the flash of the device we need to patch flashrom as follows:

git clone https://github.com/flashrom/flashrom
cd flashrom/
git checkout c8305e1dee66cd69bd8fca38bff2c8bf32924306
patch -p0 < ../fernly/flashrom-fernvale.patch
# manually fix Makefile.rej to complete patching

The patch does not cleanly apply so you need to fix the rejected Makefile (Makefile.rej) manually yourself.

Once this was done we can first load the fernly firmware into the devices RAM via:

fernly/build/fernly-usb-loader -w /dev/fernvale fernly/build/stage1.bin fernly/build/firmware.bin

Next, we can use the fernvale_spi programmer we patched into flashrom.

We first let it recognize the flash via:

flashrom/flashrom --programmer fernvale_spi:dev=/dev/fernvale

And then read the flash via:

flashrom/flashrom --programmer fernvale_spi:dev=/dev/fernvale -c "MX25L3205(A)" --read flash.dat

flash.dat will now contain the devices flash memory.

Writing flash attempt

Writing the flash can be performed via:

flashrom/flashrom --programmer fernvale_spi:dev=/dev/fernvale -c "MX25L3205(A)" --write flash.dat

However, the flash seems to be block protected and the block protect bits can not be disabled by flashrom. I have not (yet) found a way to disable the block protect.

Analysis

Mostly for my personal education I did some more analysis then the obligatory firmware dump.

SIM sniffing (via SIMtrace)

First, I sniffed the communication between the device and the SIM. Interestingly, it accessed all records of the telephone book and SMS storage. More specifically it accesses the following files, which are not needed to provide the services rendered by the device itself:

• ADF
  • EF(ECC)
  • EF(EXT2)
  • EF(SMS)
• DF(TELECOM)
  • DF(PHONEBOOK)
    • EF(ADN)
    • EF(ANRA1)
    • EF(SMS)

Other SIM accesses seems to be normal.

This is probably not an elaborate scheme to harvest phone numbers and send them to China, but rather the way the default manufactured SIM code was implemented and it was never trimmed down to the needs of this device. Nevertheless, I found it interesting seeing how the device is accessing virtually everything on the SIM.

GPRS sniffing attempt (via OpenBTS)

Next, I tried to sniff the Internet traffic to figure out what is send to whom via the mobile data connection. To this end, I used a Ettus B100 with OpenBTS.

Unfortunately, the S8 data line locator did not connect to the GPRS. This caused the following alternative response to the dw location command:

Loc:Please link:http://gpsui.net/smap.php?lac=1000&cellid=10&c=901&n=70&v=7100 Battery:67%

Flash contents

The most interesting things could be found in the dumped flash.

OS

Strings in the flash.dat suggest the device is probably running Nucleus RTOS:

$ strings -a flash.dat
Copyright (c) - ATI - Nucleus PLUS - Version ARM /  1.11.

Other strings that may help identify the OS are:

$ strings -a flash.dat | grep "\.c"
psss\components\src\bl_Secure_v5.c
psss\components\src\SSS_secure_shared_v5.c
hal\system\bootloader\src\bl_Main.c
hal\system\bootloader\src\bl_Main.c
hal\system\bootloader\src\bl_FTL.c
hal\system\bootloader\src\bl_FTL.c
hal\system\bootloader\src\bl_FTL.c
hal\storage\flash\mtd\src\flash_disk.c
hal\system\bootloader\src\bl_Main.c
hal\peripheral\src\dcl_pmu6261.c
hal\system\cache\src\cache.c
hal\peripheral\src\dcl_rtc.c
hal\peripheral\src\dcl_pmu6261.c
hal\system\bootloader\src\bl_FTL.c
hal\system\bootloader\src\bl_FTL.c
hal\peripheral\src\rtc.c
hal\peripheral\src\rtc.c
hal\peripheral\src\rtc.c
hal\peripheral\src\rtc.c
hal\peripheral\src\rtc.c
hal\peripheral\src\rtc.c
hal\storage\flash\mtd\src\flash_mtd_sf_dal.c
hal\peripheral\src\dcl_pmu_common.c
hal\peripheral\src\dcl_f32k_clk.c
hal\peripheral\src\dcl_f32k_clk.c
hal\peripheral\src\dcl_gpio.c
hal\peripheral\src\dcl_pmu_common.c
hal\system\cache\src\cache.c
hal\peripheral\src\dcl_f32k_clk.c
hal\peripheral\src\dcl_gpio.c
hal\peripheral\src\gpio.c
hal\system\bootloader\src\bl_FTL.c
hal\peripheral\src\rtc.c
hal\peripheral\src\bmt_hw.c
hal\peripheral\src\dcl_pmu6261.c
hal\storage\flash\mtd\src\flash_mtd.c
hal\peripheral\src\gpio.c
custom\common\hal\combo_flash_nor.c
hal\peripheral\src\dcl_rtc.c
hal\peripheral\src\dcl_rtc.c
hal\storage\flash\mtd\src\flash_disk.c
custom\common\hal\combo_flash_nor.c
hal\storage\flash\mtd\src\flash_mtd_sf_dal.c
hal\system\emi\src\emi.c
sss\components\src\SSS_secure_shared_common.c
alice.c
ddload.c
plutommi\Framework\GDI\gdisrc\gdi.c
C.cKi
hal\audio\src\v1\audio_service.c
ddload.c
ddload.c
plutommi\Framework\GDI\gdisrc\gdi_image_hwjpg_v2.c
plutommi\Framework\GDI\gdisrc\gdi_image_hwjpg_v2.c
plutommi\Framework\GDI\gdisrc\gdi_util.c
plutommi\Framework\GDI\gdisrc\gdi_util.c
hal\audio\src\v1\audio_service.c
ddload.c

FAT12 filesystems (?)

Searching the flash.dat for the FAT12 file systems that are supposedly present in on MediaTek phones, we get:

$ hexdump -C flash.dat
002c1e20            ef cd 4e 4f  4e   |...........NO NA|
002c1e30  4d                  |ME    FAT12   ..|
002c1e40                    |................|
002c1e50             4d 4d 4d 4d 4d 4d  |..........MMMMMM|
002c1e60  4d 4d 4d 4d 4d 4d 4d 4d  4d 4d 4d 4d 4d 4d 4d 4d  |MMMMMMMMMMMMMMMM|
*
002c1ff0  4d 4d 4d 4d 4d 4d 4d 4d  4d 4d 4d 4d 4d 4d  aa  |MMMMMMMMMMMMMMU.|
[...]
002d8400  eb     6c             |.X.FileSys .....|
002d8410     9b  f8             |................|
002d8420  9b            4e 4f  4e   |......)..!0NO NA|
002d8430  4d                  |ME    FAT12   ..|
002d8440                    |................|
002d8450             4d 4d 4d 4d 4d 4d  |..........MMMMMM|
002d8460  4d 4d 4d 4d 4d 4d 4d 4d  4d 4d 4d 4d 4d 4d 4d 4d  |MMMMMMMMMMMMMMMM|
*
002d85f0  4d 4d 4d 4d 4d 4d 4d 4d  4d 4d 4d 4d 4d 4d  aa  |MMMMMMMMMMMMMMU.|
[...]
002dbc00  ff ff ff ff ff de 9e         ba ff  |.......h.....P..|
002dbc10                    |U...............|
002dbc20            ef cd 4e 4f  4e   |...........NO NA|
002dbc30  4d                  |ME    FAT12   ..|
002dbc40                    |................|
002dbc50             4d 4d 4d 4d 4d 4d  |..........MMMMMM|
002dbc60  4d 4d 4d 4d 4d 4d 4d 4d  4d 4d 4d 4d 4d 4d 4d 4d  |MMMMMMMMMMMMMMMM|
*
002dbdf0  4d 4d 4d 4d 4d 4d 4d 4d  4d 4d 4d 4d 4d 4d  aa  |MMMMMMMMMMMMMMU.|

However, two of the partitions do not appear to be valid FAT12 file systems:

$ fls -o  flash.dat -f fat12
Invalid magic value (Not a FATFS file system (magic))
$ fls -o  flash.dat -f fat12
v/v :   $MBR
v/v :   $FAT1
v/v :   $FAT2
d/d :   $OrphanFiles
$ fls -o  flash.dat -f fat12
Invalid magic value (Not a FATFS file system (magic))

And the middle FAT12 block seems to be corrupted as well, i.e. only orphan files:

$ fls -o  flash.dat -rp -f fat12
v/v :   $MBR
v/v :   $FAT1
v/v :   $FAT2
d/d :   $OrphanFiles
-/r * :  $OrphanFiles/MP0B_001
-/r * :  $OrphanFiles/ST33A004
-/r * :  $OrphanFiles/ST33B004
[...]

An attempt was made to extract the files:

fls -o  flash.dat -Frp -f fat12 | while read line; do
    path=$(echo "$line" | awk -F':' '{print $2}')
    mkdir -p $(dirname $path);
    icat -o  flash.dat $(echo "$line" | grep -oE "[0-9]+" | head -n1) > $path
done

But most files are empty. The results are also very inconsistent, i.e., when changing SIM cards there are significant changes to the files listed by The Sleuthkit. This indicates that those are either not FAT12 partitions or a modified FAT12 variant.

Further analysis was hence done using hexdump.

Configuration data

The flash also contained some configuration data. First, the IMSI of the inserted SIM and the number that is used to remote control the device could be found in the flash:

$ hexdump -C flash.dat
002e2ad0       xx xx xx  xx xx xx xx xx xx xx xx  |....2xxxxxxxxxxx|
002e2ae0  xx xx                 |xx7.............|
002e2af0                    |................|
*
002e2b10              2b    xx  |...........+491x|
002e2b20  xx xx xx xx xx xx xx xx           |xxxxxxxx........|
002e2b30                    |.............gps|
002e2b40    2e 6e               |ui.net..........|
002e2b50                    |.............gps|
002e2b60    2e 6e               |ui.net..........|
002e2b70                    |................|
*

In the above flash segment you can also see a reference to gpsui.net. This is presumably the remote server which is contacted to turn the MCC, MNC, LAI and CID codes into street, city and country information as well as the link to gpsui.net which forwards to Google maps. However, because writing to the flash could not be achieved this hypothesis could not be confirmed.

Hidden commands

Eventually, there was a small find potentially making this effort worthwhile. Searching the flash.dat for the dw,1111 and 0000 commands reveals more hidden commands:

$ hexdump -C flash.dat
  8c ae   8d ae 8e ae      6f     |........restore.|
                    |............hhh.|
                    |................|
   6d                 |imsi............|
        6d   7a 6f 6e       |....timezone....|
             6d       |........time....|
                    |............aqe.|
000695a0                    |................|
000695b0                    |aqc.............|
000695c0                    |....server......|
000695d0                    |........ddd.....|
000695e0                    |............reg.|
000695f0               0a     |................|
                    |aqb.............|
  0b                  |....qqq.........|
      0c              |........dw......|
           0d    6c 6f    |............loc.|
               0e     |................|
                    |faa.............|
  0f                  |....fff.........|
                    |............|
                    |............|
                    |................|
000696a0                    |rpt.............|
000696b0                    |....gbrpt.......|
000696c0               6b     |........track...|
000696d0               6d 6f 6e   |............moni|
000696e0   6f                 |tor.............|
000696f0   6f  6f 6e              |soson...........|
       6f  6f           |....sosoff......|
            6f        |........sos.....|
                  6f  |............qcso|
               1a     |s...............|
  6c   6f 6e              |ledon...........|
  1b    6c   6f           |....ledoff......|
      1c      6c     6f 6e  |........flighton|
           1d     6c    |............flig|
    6f           1e     |htoff...........|
     6f 6e 6f             |esionow.........|
000697a0  1f       6f           |....esioaddr....|
000697b0              6f 6e     |.... ...hbton...|
000697c0                  6f  |........!...hbto|
000697d0                    |ff.........."...|
000697e0     6f 6c 6f             |esiolocatetype..|
000697f0                    |#...eee.........|
            6e    6f    |....$...sndstop.|
                    |........%...dde.|
                    |............&...|
   6f  6d               |formattf........|
        6c            |'...help........|
           2a    2a     |....(...*e81*...|
               2a     |........)...*e80|
  2a             2a     |*...........*...|
  2a    6f 6f  2a           |*reboot*........|
  2b    2a         2a     |+...**...|
000698a0      2c      6d        |....,...imeiset.|

However, most of those commands do not function correctly. It seems the devices firmware is shared among several such location tracking and listening devices, e.g., there are commands referring to LEDs and a TF card, both of which this device do no feature, however, other devices available online do.

An incomplete list of the found commands and there replies is:

• help: replies with the following commands:
  • dw: Locate
  • qqq: Device binding
  • 1111: Sound Alarm Monitor on
  • 0000: Sound Alarm Monitor off
  • ddd: Reset all tasks
  • aqb: Get Username Password
  • eee: Recording saved
  • dde: Cleanup TF card
  • hhh: Device status
• loc: same as dw
• imsi: Query IMEI and IMSI
• faa: “DTMG: Set voice monitoring, SMS reply and sound sensitivity successfully:40”, “DTMG: Unusual sound detected”
• fff: “DT: Set voice monitoring, voice callback and sound sensitivity successfully:40”
• 1111: “DT: Set voice monitoring, voice callback and sound sensitivity successfully:400”
• 0000: “DT: Voice monitoring cancelled successfully.”
• gbrpt: “Report:Location the continuous escalation has been closed.”
• track: “Track:Caller answer mode the device is set to reply location.”
• hbton: “Hbt:Device is turned on real-time online”
• hbtoff: “Hbt: Device online has been closed”
• esionow: “…” ?
• esioaddr: “Setting esio addr and port fail!”
• esiolocatetype: “Esio:Reporting location type has been updated to 0.”
• server: “Setting server addr and port fail!”
• reg: “…” ?
• monitor: “Monitor:Caller answer mode the device is set to automatically answer.”
• eee: “Tf-Card check fails of is insufficient free space!”
• sndstop: “Cam:No task is running, cancel failed!”
• *e81*: “…” ?
• *e80*: “…” ?
• soson, sosoff, sos, qcsos: ?
• ledon, ledoff: ?
• flighton, flightoff: ?
• aqe: “Setting apn fail!”
• imeiset: “…” does not seem to set the IMEI
• restore: “Restore ok!”
• formattf: ?
• time: “…” ?
• timezone: “Setting time zone ok. Current time zone 0”
• age: “…” ?
• *3646655*: queries for version information
• *reboot*: reboots the device

Interestingly the reply strings could not be found in the flash in plaintext. This suggests that some of the data is compressed.

The message log of me trying some of the found hidden commands to populate the above list is as follows:

It seems that we can use esioaddr to change the address used to lock up the location information. However, no connection to a given domain nor IP is actually made. The device will simply report the addr invalid in the location report.

The server command sets a different server. Changing it does not result in the addr invalid responses, as can be seen from this second message log:

Provider call logs and itemized bill

Because the GPRS sniffing failed I resort to the billing of my provider to further analyze the communication habits of the S8 data line locator.

Obviously reply SMS are billed. More interesting is the Internet access patterns.

dw or loc commands and during idle

During location queries the device will use “MMS/Internet” service. The following is a segment in which first repeated location queries were performed, then the devices was left idle:

Even during idle the device sometimes uses the “MMS/Internet” service.

Even though I deactivated all tracking features that I may have activated during my previous testing, I can not be 100 % sure that this is not something that I activated, maybe while stumbling through the gpsui.net website. However, I regardless of whether I activated this “feature” or not, I do not want it and would like to know what data is actually send and how to deactivate it.

gpsui.net

Going deeper into the gpsui.net website would probably result in a new writeup in itself. It is a very big surveillance hub, just replace the xxxx in http://gpsui.net/u/xxxx with some letters and numbers and you can see random people’s locations.

The website also makes a reference to ZhiPu:

It seems this is the company that makes these trackers.

You can get your credentials for login by texting aqb to your S8 data line monitor. The username as well as the password are 6 digit numbers. They are also located in the flash right before the IMSI.

Detection

When sending data the S8 data line locator can be detected with a CC308+ (a cheap Chinese RF detector):

The S8 data line locator seems to be badly shielded. A location request via the dw command causes noticeable electronic noise by the device. It seems in general to cause all sorts of RF interference.

Future work

While I did not (yet) succeed with my original goal to disable the mobile data phone home “feature’, it was nevertheless a fun exercise and hopefully someone finds this useful or at least educational.

Future work needs to be done on several things:

Issues

I was not able yet to write new firmware via flashrom because I was not able to disable block protection on the flash, yet. Maybe a different avenue for flashing new firmware could be the SPFlash tool4 and/or the Flash tool. However, that would not be open source. If know something about the weird FAT12 file system used in the device or are able to flash your S8 data line locator please contact me with details!

Further, I tried to capture the GPRS data connection of the device, but was unable to do so. It would not use GPRS when connected to my network. Currently, I do not know how the APN is configured. The SIM trace does not indicate that the EF(ACL) is ever accessed. However, as I found the correct APN configuration stored in the devices flash, this suggests the device acquires this information via a setup SMS by the service provider.

Ideas

Dremel the board smaller, e.g., you don’t need the USB connector. This way the S8 data line locator could be turned into a “modular” bug that could be placed where ever there is a 5 V 1 A power source.

Appendix: Fuck up

No writeup would be complete without at least one fuck up. So here it is:

While using the S8 data line locator with OpenBTS I provisioned imaginary numbers. When switching SIM cards I forgot to turn of the voice activated callback.

So long story short, some guy with the number 3333333 listend in on me for 2 minutes:

I did not notice this until I reviewed the logs!

So my resume on these little hardware espionage implants: They are stealthy and dangerous as fuck!

Bibliography

[1] NSA/CSS, “ANT product catalog (USB),” leaked documents (pdf: https://cryptome.org/2013/12/nsa-ant-usb.pdf).

[2] A. “bunnie” Huang and S. X. Cross, “Fernvale: An open hardware and software platform, based on the (nominally) closed-source mT6260 soC,” talk at the 31st Chaos Communication Congress (slides: http://www.bunniefoo.com/fernvale/fernvale-31c3.pdf, video: https://www.youtube.com/watch?v=hpEqDPYtf9s).

原文链接：https://ha.cking.ch/s8_data_line_locator/