In general, a “glitch” is a momentary fault that may happen on a system, preventing it from working properly, for a brief amount of time. The effects of a single glitch on proper software execution may be multiple, including catastrophic consequences that may prevent the system from continuing the execution.
Glitching attacks are complex and expensive to execute, but can be a real issue for secure boot mechanisms, and often very hard to prevent or mitigate. They aim at exploiting predictable consequences of single glitches in order to take control of the execution or the data contained in the system. The glitch can be injected using different techniques, which often rely on well known weaknesses of the specific microcontroller or CPU. The most common glitch injection consists in varying the voltage supplied to the chip at a specific time, or modifying the profile of the clock signal to mangle the timing of the execution of the instructions. More advanced attacks can rely on irradiating the device with strong electromagnetic interference.
In the specific context of secure boot, the goal for an attacker is to circumvent the security checks in those critical sections of the code, e.g. the code that performs verifications on the firmware authenticity, integrity or versioning. These attacks could eventually defeat the security checks, and take control of the system by uploading an unauthorized firmware image. While they require an accurate synchronization and several attempts, these attacks will eventually succeed in injecting a fault in the hardware at the required time in order to skip the verification.
Our secure bootloader, wolfBoot, follows the indication of RFC9019 to provide a secure, public key based verification of the integrity and authenticity of the firmware and its updates. It runs on several different architectures, from small microcontrollers up to x86_64 systems. wolfBoot is OS-agnostic and provides best-in-class security thanks to the FIPS 140-2 certified algorithms implemented in the wolfCrypt security engine.
wolfBoot already comes with plenty of unique features. Now it is also the first open source secure bootloader to implement mitigations against glitching attacks. Our development team has recently added an optional feature that can be activated at compile time, to reinforce the security of the critical variables and decision points in the code. This has required an evaluation of the code flow of wolfBoot from a point of view that includes the possibility for an attacker to skip single specific instructions. Introducing these mitigations has been tricky, because redundant code written in C is usually discarded by the compiler. For this reason the countermeasures must be programmed in assembly, which makes this code architecture specific.
The upcoming release of wolfBoot will contain a first version of these countermeasures, but glitching support mitigation already made it to our main branch on github, and it can be freely compiled and used in GPL projects for evaluation and auditing purposes.
To compile wolfBoot with glitching and side-channel attack mitigations turned on, it is sufficient to add ARMORED=1 to the configuration options (i.e. via command line when invoking make, or through the .config file). The ARMORED option is currently supported on ARM Cortex-M architecture. Support for other architectures will be added in the future.
What is the next feature that you want to see implemented in wolfBoot? Is there any architecture or platform that we don’t yet support that could benefit from our glitch-resistant secure boot mechanism? Let us know! Drop us a line at email@example.com.