36C3 - Open Source is Insufficient to Solve Trust Problems in Hardware

Bunnie Huang explains why open source is insufficient for hardware trust due to the destructive nature of verification, presenting the Betrusted platform.

media.ccc.de/v/36c3-10690-open_source_is_insufficient_to_sol… How Betrusted Aims to Close the Hardware TOCTOU Gap While open source is necessary for trustable hardware, it is far from sufficient. This is because “hashing” hardware – verifying its construction down to the transistor level – is typically a destructive process, so trust in hardware is a massive time-of-check/time-of-use (TOCTOU) problem. This talk helps us understand the nature of the TOCTOU problem by providing a brief overview of the supply chain security problem and various classes of hardware implants. We then shift gears to talk about ways to potentially close the TOCTOU gap, concluding with a curated set of verifiable components that we are sharing as an open source mobile communications platform – a kind of combination hardware and software distribution – that we hope can be useful for developing and deploying all manner of open platforms that require a higher level of trust and security. The inconvenient truth is that open source hardware is precisely as trustworthy as closed source hardware. The availability of design source only enables us to agree that the designer’s intent can be trusted and is likely correct, but there is no essential link between the hardware design source and the piece of hardware on your desk. Thus while open source is necessary for trustable hardware, it is far from sufficient. This is quite opposite from the case of open source software thanks to projects like Reproducible Builds, where binaries can be loaded in-memory and cryptographically verified and independently reproduced to ensure a match to the complete and corresponding source of a particular build prior to execution, thus establishing a robust link between the executable and the source. Unfortunately, “hashing” hardware – verifying its construction down to the transistor level – is typically a destructive process, so trust in hardware is a massive time-of-check/time-of-use (TOCTOU) problem. Even if you thoroughly inspect the design source, the factory could modify the design. Even if you audit the factory, the courier delivering the hardware to your desk could insert an implant. Even if you carried the hardware from the factory to your desk, an “evil maid” could modify your machine. This creates an existential crisis for trust – how can we know our secrets are safe if the very hardware we use to compute them could be readily tainted? This talk addresses the elephant in the room by helping us understand the nature of the TOCTOU problem by providing a brief overview of the supply chain security problem and various classes of hardware implants. We then shift gears to talk about ways to potentially close the TOCTOU gap. When thinking about hardening a system against supply chain attacks, every component – from the CPU to the keyboard to the LCD – must be considered in order to defend against implanted screen grabbers and key loggers. At every level,