phd-thesis/abstract.tex

46 lines
4 KiB
TeX

\chapter*{Abstract}
\adjustmtc
\addcontentsline{toc}{chapter}{Abstract}
In the past decades, cryptographic advancements and techniques like formal verification have steadily improved software
security. Meanwhile, the field of hardware security has not kept pace. Research has made progress in subfields such as
resilience to Side-Channel Attacks (SCA) and Physical Unclonable Functions (PUFs). However, the state of the art still
often relies on microelectronic integration to achieve security by obscurity insted of more fundamental security
guarantees. While effective, system-level tamper protection is only used in few devices such as Hardware Security
Modules (HSMs) and card payment terminals. Due to the high cost and low performance of HSMs in particular, they remain
relegated to niche applications such as Transport Layer Security (TLS) certificate issuance and payment data processing.
In this thesis, we introduce the Inertial Hardware Security Module (IHSM), a new architecture for low-cost hardware
security modules that provide high-level active tamper protection, while supporting computing payloads of much larger
size, weight and power dissipation compared to conventional HSMs. In an IHSM, the costly and difficult to source
tamper-sensing mesh of a conventional HSM is replaced by a mesh made from simple PCBs that is rotating at high speed
around the payload. Since the mesh is rotating at high speed, it cannot be manipulated, and the security of conventional
meshes created in bespoke manufacturing processes can be achieved using much simpler and less expensive construction
techniques. We present the results of a survey of approximately 30 real world tamper sensing mesh implementations. Based
on our findings, we deduce design criteria for secure meshes and contextualize our design. We further motivate the
necessity of secure hardware by presenting an analysis of problematic aspects in the hardware security design of
Germany's new national electronic health record system.
To pave the way for practical implementations of IHSM technology, we present solutions to key engineering challenges in
IHSM construction. We present a design and analysis of highly symmetric planar inductors for rotating wireless power
transfer that improves self-resonant frequency by up to \qty{58}{\percent} and inductance by up to \qty{6.5}{\percent}
in our tests. Complementing this research, we present a high-fidelity, low-cost monitoring system for security meshes
that is based on the principles of Time-Domain Reflectometry (TDR), reaching \qty{184}{\pico\second} time resolution. We
validate our system and find that it is able to reliably detect several classes of advanced physical attacks. We find
that our system is sensitive enough to detect differences between identical copies of the same mesh, suggesting PUF-like
properties.
Applying IHSM technology, we analyse two use cases that are unlocked by the increased size and power dissipation
capability of IHSMs. In the first analysis, an IHSM-secured relay node for Quantum Key Distribution (QKD) systems is
proposed, enabling their practical implementation across arbitrary distances, which requires trusted relay stations due
to fundamental physical limitations. In the study, IHSMs are adapted for such high-security QKD relays by securing the
IHSM mesh passthrough with a secondary tamper-sensing mesh. In this setup, a bracket design is proposed that supports
passing through optical fibers at low loss.
The second proposed use case adapts an IHSM enclosure to the size, power and thermal dissipation requirements of a
high-power server to support co-located secure Multiparty Computation (MPC) workloads. In practical MPC deployments,
nodes are distributed across data centers to avoid a single point of failure for physical attacks. As a result,
practical MPC deployments are limited by network bandwidth and latency constraints. Using IHSMs, physically secured MPC
nodes can be deployed within the same data center, increasing bandwidth, reducing latency and unlocking a new
performance spectrum.