Engine or Platform Security Processor; it is more deeply embedded than any known coprocessor, having access to not only all of the CPU’s memory, but its register file and execution pipeline as well.
It is thought that only VIA C3 CPUs are affected by this issue. The C-series processors are marketed towards industrial automation, point-of-sale, ATM, and healthcare hardware, as well as a variety of consumer desktop and laptop computers.
The scope of this is limited; generations of CPUs after the C3 no longer contain this feature.
This work is released as a case study and thought experiment, illustrating how backdoors might arise in increasingly complex processors, and how researchers and end-users might identify such features. The tools and research offered here provide the starting point for ever-deeper processor vulnerability research.
Checking your CPU
To check if your CPU is affected:
git clone https://github.com/xoreaxeaxeax/rosenbridge
sudo modprobe msr
The provided utility must be run on baremetal (not in a virtual-machine), and is in an alpha state. It may crash, panic, or hang systems not containing the backdoor.
The utilities provided here are designed around a specific processor family and core; unfortunately, the tools will miss the backdoor if it has been even slightly modified from the researched form.
Closing the Backdoor
Some systems have the backdoor enabled by default, allowing unprivileged code to gain kernel level access without permission. If the steps in ‘Checking your CPU’ indicate that your CPU is vulnerable, you can install a script to close the backdoor early in the boot process:
sudo make install
Note that, even with this, an attacker with kernel level access can still re-enable the backdoor. This script is provided as an outline for correcting the issue during the boot process, but will require adaptation for different systems.
Tools and Techniques
The utility is used extensively in this research for uncovering unknown instructions.
An for the Deeply Embedded Instruction Set (DEIS). It converts programs written in the custom rosenbridge assembly into x86 instructions, which, when executed following the launch-instruction, will send the commands to the hidden CPU core.
A proof-of-concept of using the rosenbridge backdoor for privilege escalation.
A rough outline for closing the vulnerability on affected systems, to the extent possible through model-specific-register updates.
A collection of utilities used to fuzz both the x86 and rosenbridge cores, in order to isolate the unknown launch-instruction and bridge-instruction, and resolve the instruction format of the rosenbridge core.
The fuzzer used to explore the effects and capabilities of the hidden CPU core.
It is thought that, on some processors, an exit sequence is needed to switch back to the x86 core at the end of a DEIS sequence. This directory contains the utilities used to search for the exit sequence in early stages of the research, but was abandoned when a processor was found not requiring any such sequence.
A collection of python utilities designed to monitor and manage fuzzing tasks across a network of workers.
A stripped down version of the sandsifter fuzzer, used to identify the bridge-instruction that will send commands from the x86 core to the hidden rosenbridge core.
A collection of helper utilities used to monitor kernel memory and registers for changes caused by fuzzed DEIS instructions.
Utilities to lock or unlock the rosenbridge backdoor.
A tool to identify patterns from the fuzzing logs to identify classes of DEIS instruction behaviors.
A tool used early in the research, to attempt to identify the hidden core’s architecture by executing known RISC instructions.
An alpha-state tool to detect whether or not a processor is affected by rosenbridge.
The details and implications presented in this work are the authors’ inferences and opinions, derived from the research described. The research is performed and provided with the goal of identifying and fixing a perceived security vulnerability on the described CPUs. VIA processors are renowned for their low power usage and excellence in embedded designs; we believe that the functionality described was created in good faith as a useful feature for the embedded market, and was unintentionally left enabled on some early generations of the processor. No malicious intent is implied.
project:rosenbridge is a research effort from Christopher Domas (@xoreaxeaxeax).