When it comes to large systems, there are a lot more computers than there are people maintaining them. That’s not a big deal since you can simply use a KVM to connect one Keyboard/Video/Mouse terminal up to all of them, switching between each box simply and seamlessly. The side effect is that now the KVM has just as much access to all of those systems as the human who caresses the keyboard. [Yaniv Balmas] and [Lior Oppenheim] spent some time reverse engineering the firmware for one of these devices and demonstrated how shady firmware can pwn these systems, even when some of the systems themselves are air-gapped from the Internet. This was their first DEF CON talk and they did a great job of explaining what it took to hack these devices.
[Craig Heffner] has been busy with his Linksys WRT120N router. When we last checked in on [Craig] he had reverse engineered the obfuscation techniques used in the router’s firmware. Since then, he’s re-enabled JTAG, cracked the “encryption” used for saving configuration backups, and now he’s devised a simple attack to change the admin password. With the firmware unlocked, [Craig] went after the hardware JTAG. His first hurdle was a missing jumper connecting the TDI pin to the processor. With a solder blob making the connection, he then found the router would connect to his JTAG debugger, and immediately reset. TDI had been re-used as a GPIO in software, and assigned to the reset button on the back of the router. [Craig’s] JTAG pod was pulling the pin low and causing the reset. To make matters worse, the bootloader also redefined and checked for the reset button. If the button were pressed it would boot into a recovery mode. [Craig] patched the bootloader with a little help from IDA pro. He then desoldered the router’s flash and programmed it outside the system. The firmware required a similar patch. Rather than desolder the flash chip again, [Craig] created a firmware update the router would accept and flashed it via the router’s web interface.
Since he already was deep into the Linksys Firmware, [Craig] looked for any obvious attack vectors. He found a big one in the /cgi/tmUnBlock.cgi. Inside the firmware, the URL sent to the CGI would be sent through sprintf(). In plain english, it means that no input length checking was happening – so a URL longer than the firmware engineers expected (in this case 256 bytes) would overflow into areas of memory it wasn’t supposed to – in this case, the stack. For an astute attacker, that’s a wide open door. [Craig] was able to use find some Return Oriented Programming (ROP) gadgets and created an input value that would cause the router to reset its own administrator password. After running the exploit, a quick trip to the router’s webpage proved his attack was successful.
If that wasn’t enough, [Craig] also spent some time looking at the patches to the router’s firmware. The release notes of one of the patches mentioned encrypting configuration files. The WRT120N, like many routers, allows the owner to download and save the configuration as a file. It turned out that the “encryption” scheme was nothing more than an exclusive OR with 0xFF. A pretty weak encryption scheme by any standards. To [Craig] we send our congratulations. To the WRT120N software engineers, we’d suggest taking one of [Craig’s] embedded device exploitation classes.
[Craig Heffner] recently found himself on the case of the Linksys WRT120N router. The router’s firmware was using some previously unknown form of obfuscation, causing headaches for those wishing to run their own software. The WRT120N, being a 2009 model is somewhat out of date at this point. That didn’t stop [Craig] though, as he dove into reverse engineering the firmware obfuscation.
[Craig] started by running the firmware through his own Binwalk tool. Binwalk analyzes firmware files for known data, be it embedded filesystems, raw compression streams, or binary files. In this case Binwalk only found a small LZMA block which contained the compressed html files for the router’s web interface. The rest of the firmware was unknown data with a high level of entropy. [Craig] couldn’t do anything more with the firmware update file alone, so he ordered a router to attack from the hardware side. Inside he found typical low-end router components: An Atheros AR7240 SoC, a 2MB SPI flash chip, 32MB of RAM. He also found serial and JTAG headers.
[Craig] connected to the serial port and was greeted with a boot menu. This allowed him to run some commands on the router, but didn’t give him any way to dump memory. He had to go straight to the source – connecting directly to the router’s SPI flash with an FTDI C232HM cable. Using libmpsse, another of his open source tools, [Craig] was able to dump the flash. He now had the un-obfuscated bootloader code, albeit in MIPS assembly. [Craig] was then able to go after the bootloader with IDA Pro. After a bit of work, the obfuscation system was exposed. The system was simple – several byte and nibble swaps had been performed between the LZMA header block and the first few bytes of data. [Craig] finished out this part of his hack by writing a simple C program to de-obfuscate and decompress the firmware.