jaseg/ihsm
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Improve section transitions

Browse source
This commit is contained in:
jaseg 2021-01-05 17:30:08 +01:00
parent 1bd8ebc041
commit 92f646fbf3
1 changed files with 112 additions and 96 deletions
Download patch file Download diff file Expand all files Collapse all files

208
doc/quick-tech-report/rotohsm_tech_report.tex
View file

@ -146,6 +146,9 @@ Section~\ref{sec_conclusion}.
% summaries of research papers on HSMs. I have not found any actual prior art on anything involving mechanical motion
% beyond ultrasound.
In this section, we will briefly explore the history of HSMs and the state of academic research on active tamper
detection.
HSMs are an old technology tracing back decades in their electronic realization. Today's common approach of monitoring
meandering electrical traces on a fragile foil that is wrapped around the HSM essentially transforms the security
problem into the challenge to manufacture very fine electrical traces on a flexible foil~\cite{isaacs2013, immler2019,
@ -188,7 +191,7 @@ compound that has been loaded with RF-reflective grains. In their concept, the R
transceivers is shaped by the precise three-dimensional distribution of RF-reflective grains within the potting
compound.
Our concept is novel in that mechanical motion has not been proposed before as part of a hardware security module. Most
We are the the first to propose a mechanically moving HSM security barrier as part of a hardware security module. Most
academic research concentrates on the issue of creating new, more sensitive security barriers for HSMs~\cite{immler2019}
while commercial vendors concentrate on means to cheaply manufacture and certify these security
barriers~\cite{drimer2008}. Our concept instead focuses on the issue of taking any existing, cheap low-performance
@ -244,38 +247,40 @@ necessary.
\subsection{Mechanical layout}
The simplest way to mount a stationary payload in a spinning security mesh is to use a hollow shaft. The payload can be
mounted on a fixed rod threaded through this hollow shaft along with wires for power and data. The shaft is a weak spot
of the system, but this weak spot can be alleviated through either careful construction or a second layer of rotating
meshes with a different axis of rotation. Configurations that do not use a hollow-shaft motor are possible, but may
require additional bearings to keep the stator from vibrating.
Thinking about the concrete construction of our mechanical HSM, the first challenge is mounting both mesh and payload on
a single shaft. The simplest way we found to mount a stationary payload inside of a spinning security mesh is a hollow
shaft. The payload can be mounted on a fixed rod threaded through this hollow shaft along with wires for power and
data. The shaft is a weak spot of the system, but this weak spot can be alleviated through either careful construction
or a second layer of rotating meshes with a different axis of rotation. Configurations that do not use a hollow-shaft
motor are possible, but may require additional bearings to keep the stator from vibrating.
The spinning mesh must be designed to cover the entire surface of the payload during one revolution. Still, it can be
designed with longitudinal gaps to allow outside air to flow through to the payload. In boundary-sensing HSMs, cooling
of the processor inside is a serious issue since any air duct or heat pipe would have to penetrate the HSM's security
boundary. This problem can only be solved with complex and costly siphon-style constructions, so in commercial systems
heat conduction is used exclusively~\cite{isaacs2013}. This limits the maximum power dissipation of the payload and thus
its processing power. Our setup allows direct air cooling, which increases the maximum possible power dissipation of
the payload and unlocks much more powerful processing capabilities. Instead of gaps one could even integrate an actual
fan into the rotor.
The next design choice we have to make is the physical structure of the security mesh. The spinning mesh must be
designed to cover the entire surface of the payload, but compared to a traditional HSM it suffices if it sweeps over
every part of the payload once per rotation. This means we can design longitudinal gaps into the mesh that allow outside
air to flow through to the payload. In traditional boundary-sensing HSMs, cooling of the payload processor is a serious
issue since any air duct or heat pipe would have to penetrate the HSM's security boundary. This problem can only be
solved with complex and costly siphon-style constructions, so in commercial systems heat conduction is used
exclusively~\cite{isaacs2013}. This limits the maximum power dissipation of the payload and thus its processing power.
Our setup allows direct air cooling of regular heatsinks. This greatly increases the maximum possible power dissipation
of the payload and unlocks much more powerful processing capabilities. In an evolution of our design, the spinning mesh
could even be designed to *be* a cooling fan.
\subsection{Spinning mesh power and data transmission}
The basic concept of a security mesh spinning at more than $\SI{500}{rpm}$ around a payload leaves us with a few
On the electrical side, the idea of a security mesh spinning at more than $\SI{500}{rpm}$ leaves us with a few
implementation challenges. Since the spinning mesh must be monitored for breaks or short circuits continuously, we need
both a power supply for the spinning monitoring circuit and a data link back to the stator.
both a power supply for the spinning monitoring circuit and a data link to the stator.
A good starting point for power transfer is a simple setup of a stationary bright lamp shining at a rotating solar
panel. In contrast to e.g.\ slip rings, this setup is mechanically durable at high speeds and it also provides
reasonable output power (see Appendix \ref{sec_energy_calculations} for some calculations on power consumption). A
battery may not provide a useful lifetime without power-optimization. Likewise, an energy harvesting setup may not
provide enough current to supply peak demand.
We found that a bright lamp shining at a rotating solar panel is a good starting point. In contrast to e.g.\ slip
rings, this setup is mechanically durable at high speeds and it also provides reasonable output power (see Appendix
\ref{sec_energy_calculations} for some calculations on power consumption). A battery may not provide a useful lifetime
without power-optimization. Likewise, an energy harvesting setup may not provide enough current to supply peak demand.
Since the monitoring circuit uses little current, power transfer efficiency is not important. On the other hand, cost
may be a concern in a production device. Here it may prove worthwhile to replace the solar cell setup with an extra
windings on the rotor of the BLDC motor driving the spinning mesh. This rotor is likely to be a custom part, so adding
these windings is unlikely to increase cost significantly. Inductive power transfer may also be an option given that one
can integrate it into the mechanical design.
winding on the rotor of the BLDC motor driving the spinning mesh. This rotor is likely to be a custom part, so adding
an extra winding is unlikely to increase cost significantly. More traditional inductive power transfer may also be an
option if it can be integrated into the mechanical design.
Besides power, the data link between spinning mesh and payload is critical to the HSM's design. This link is used to
transmit the occassional status report along with a low-latency alarm trigger (``heartbeat'') signal from mesh to payload.
@ -286,76 +291,79 @@ purpose.
\label{sec_attacks}
After outlining the basic mechanical design of an inertial HSM above, in this section we will detail possible ways to
attack it. Fundamentally, attacks on an inertial HSM are the same as those on a traditional HSM, since the tamper
detection mesh is the same. Only in the inertial HSM any attack on the mesh has to be carried out while the mesh is
rotating, which for most types of attack will require a CNC attack robot moving in sync with it. In comparison to
traditional designs, the data link between mesh and payload is an additional weak spot in the rotating desing. If it is
optical, non-contact attacks are possible.
attack it. Fundamentally, attacks on an inertial HSM are the same as those on a traditional HSM since the tamper
detection mesh is the same. Only, in the inertial HSM any attack on the mesh has to be carried out while the mesh is
rotating, which for most types of attack will require some kind of CNC attack robot moving in sync with it.
\subsection{Attacks on the mesh}
There are two locations where one can attack a tamper-detection mesh. On one hand, the mesh itself can be tampered with.
This includes bridging its traces to allow for a hole to be cut. The other option is to tamper with the monitoring
circuit itself, to prevent a damaged mesh from triggering an alarm and causing the HSM to erase its
circuit itself to prevent a damaged mesh from triggering an alarm and causing the HSM to erase its
contents~\cite{dexter2015}. Attacks in both locations are electronic attacks, i.e. they require electrical contact to
parts of the circuit. Traditionally, this contact is made by soldering or by placing a probe such as a thin needle. We
consider this contact infeasible to be performed on an object spinning at high speed without a complex setup that
rotates along with the object or that involves ion beams, electron beams or liquids. Thus, we consider them to be
practically infeasible outside of a well-funded, special-purpose laboratory.
\subsection{Attacks on the rotation sensor}
Instead of attacking the mesh in motion, an attacker may also try to first stop the rotor. To succeed, they would need
to fool the rotor's MEMS accelerometer. An electronic attack on the sensor or the monitoring microcontroller would be no
easier than directly bridging the mesh traces.
MEMS accelerometers usually use a cantilever design, where a proof mass moves a cantilever whose precise position can be
measured electronically. A topic of recent academic interest have been acoustic attacks tampering with these
mechanics~\cite{trippel2017}. In the authors' estimate these attacks are too hard to control to be practically useful
against an inertial HSM.
A possible way to attack the accelerometer inside an inertial HSM may be to first decapsulate it using laser ablation
synchronized with the device's rotation. Then, a fast-setting glue such as a cyanoacrylate could be deposited on the
moving MEMS parts, locking them in place. To mitigate this type of attack the accelerometer should be mounted in a
shielded place inside the security envelope. Further, this attack can only work if the rate of rotation and thus the
expected accelerometer readings are constant. If the rate of rotation is set to vary over time this type of attack is
quickly detected. In Appendix \ref{sec_degrees_of_freedom} we outline the constraints on sensor placement.
\subsection{Attacks on the alarm circuitry}
An electronic attack could also target the alarm circuitry inside the stationary payload, or the communication link
between rotor and payload. The link can easily be proofed by using a cryptographically secured protocol along with a
high-frequency heartbeat message. The alarm circuitry has to be designed such that it is entirely contained within the
HSM's security envelope and has to tolerate environmental attacks such as ones using temperature, ionizing radiation,
lasers, supply voltage variations, ultrasound or other vibration and gases or liquids. The easiest way to proof an alarm
system against these is to employ adequate filtering of the incoming power supply and use sensors for the others,
triggering an alarm in case extraordinary environmental variations are detected.
Besides trying to deactivate the tamper detection mesh, an electronic attack could also target the alarm circuitry
inside the stationary payload, or the communication link between rotor and payload. The link can be secured using a
cryptographically secured protocol like one would use for wireless radio links along with a high-frequency heartbeat
message. The alarm circuitry has to be designed such that it is entirely contained within the HSM's security envelope.
Like in conventional HSMs it has to be built to either tolerate or detect environmental attacks such as ones using
temperature, ionizing radiation, lasers, supply voltage variations, ultrasound or other vibration and gases or liquids.
Conventionally, incoming power rails are filtered thoroughly to prevent electrical attacks and other types of attacks
are prevented by sensors that thrigger an alarm.
If the alarm link between rotor and stator uses a spoofable interface such as an optical link, this link must be
cryptographically verified. It also must be bidirectional to allow the alarm signal receiver to verify link latency. In
a purely unidirectional spoofable link, an attacker could record the authenticated ``no alarm'' signal from the
transmitter while simultaneously replaying it just slightly slower (say at $\SI{99}{\percent}$ speed) to the receiver.
The receiver would not be able to distinguish between this attack and ordinary deviations in the transmitter's local
clock frequency. However, the attacker can at any point simply stop the rotor and replay the leftover recorded ``no
alarm'' signal. Given the frequency stability of commercial crystals, this would allow for an attack duration of several
seconds per hour of recording time.
In an inertial HSM, the mesh monitoring circuit's tamper alarm is transmitted from rotor to stator through a wireless
link. Since an attacker may wirelessly spoof this link, it must be cryptographically secured. It also must be
bidirectional to allow the alarm signal receiver to verify link latency: If it were unidirectional, an attacker could
act as a Man-in-the-Middle and replay the mesh's authenticated ``no alarm'' signal at slightly below real-time speed
(say at $\SI{99}{\percent}$ speed). The receiver would not be able to distinguish between this attack and ordinary
deviations in the transmitter's local clock frequency. Thus, after some time the attacker can simply stop the rotor and
break the mesh while replaying the leftover recorded ``no alarm'' signal. Given the frequency stability of commercial
crystals, this would yield the attacker several seconds of undisturbed attack time per hour of recording time.
\subsection{Fast and violent attacks}
A variation of the above attacks on the alarm circuitry would be an attack that
attempts to simply destroy this circuitry before the alarm can be acted upon using a tool like a large hammer or a gun.
Mitigations for this type of attack include potting the payload inside a mechanically robust enclosure. The alarm
signalling chain's integrity can be checked continuously using a cryptographic heartbeat protocol. A simple active-high
or active-low alarm signal cannot be considered fail-safe in this scenario.
\subsection{Attacks on the rotation sensor}
An attacker may try to stop the rotor before tampering with the mesh. To succeed, they would need to fool the rotor's
MEMS accelerometer. An electronic attack on the sensor or the monitoring microcontroller would be no easier than
directly bridging the mesh traces. Physical attacks on the accelerometer are possible~\cite{trippel2017}, but in the
authors' estimate are too hard to control to be practically useful.
A last type of attack might be to try to physically tamper with the accelerometer's sensing mechanism. MEMS
accelerometers usually use a cantilever design, where a proof mass moves a cantilever whose precise position can be
measured electronically. A possible way to attack such a device might be to first decapsulate it using laser ablation
synchronized with the device's rotation. Then, a fast-setting glue such as a cyanoacrylate could be deposited on the
moving MEMS parts, locking them in place. This attack would require direct access to the accelerometer from the outside
and can be prevented by mounting the accelerometer in a shielded place inside the security envelope. This attack can
only work if the rate of rotation and thus the accelerometer's readings are constant. If the rate of rotation is set to
change on a schedule, this type of attack can be detected easily. In Appendix \ref{sec_degrees_of_freedom} we outline
the constraints on sensor placement.
A variation of the above attacks on the alarm circuitry is to simply destroy the part of the HSM that erases data in
response to tampering before it can finish its job. This attack could use a tool such as a large hammer or a gun.
Mitigations for this type of attack include potting the payload inside a mechanically robust enclosure. Additionally,
the integrity of the entire alarm signalling chain can be checked continuously using a cryptographic heartbeat protocol.
A simple active-high or active-low alarm signal as it is used in traditional HSMs cannot be considered fail-safe in this
scenario as such an attack may well short-circuit or break PCB traces.
\section{Prototype implementation}
\label{sec_proto}
To validate our theoretical design, we implemented a prototype rotary HSM. The main engineering challenges we solved in
our prototype are:
After elaborating the design principles of inertial HSMs and researching potential attack vectors we have validated
these theoretical studies by implementing a prototype rotary HSM. The main engineering challenges we solved in our
prototype are:
\begin{enumerate}
\item Fundamental mechanical design suitable for rapid prototyping that can withstand a rotation of $\SI{500}{rpm}$.
\item Automatic generation of security mesh PCB layouts for quick adaption to new form factors.
\item Non-contact power transmission to rotor.
\item Non-contact power transmission from stator to rotor.
\item Non-contact bidirectional data communication between stator and rotor.
\end{enumerate}
@ -384,16 +392,18 @@ To mount the entire HSM, we chose to use ``2020'' modular aluminium profile.
\subsection{PCB security mesh generation}
To allow a quick iteration of our design while producing results with a realistic level of security, we wrote a plugin
for the KiCAD EDA suite that automatically generates parametrized security meshes. When KiCAD is used in conjunction
with FreeCAD through FreeCAD's KiCAD StepUp plugin, this ends up in an efficient toolchain from mechanical CAD design to
security mesh PCB gerber files. The mesh generation plugin can be found at its
website\footnote{\url{https://blog.jaseg.de/posts/kicad-mesh-plugin/}}.
The security mesh covers a total of five interlocking PCBs. A sixth PCB contains the monitoring circuit and connects to
these mesh PCBs. To allow us to quickly iterate our design without manually re-routing several large security meshes
for every mechanical chage we wrote a plugin for the KiCAD EDA suite that automatically generates parametrized security
meshes. When KiCAD is used in conjunction with FreeCAD through FreeCAD's KiCAD StepUp plugin, this ends up in an
efficient toolchain from mechanical CAD design to security mesh PCB gerber files. The mesh generation plugin can be
found at its website\footnote{\url{https://blog.jaseg.de/posts/kicad-mesh-plugin/}}. The meshes it produces have a
practical level of security in our application.
Our mesh generation plugin overlays a grid on the target area and then produces a randomized tree covering this grid.
The individual mesh traces are then traced along a depth-first search through this tree. A visualization of the steps is
shown in Figure \ref{mesh_gen_viz}. A sample of the production results from our prototype is shown in Figure
\ref{mesh_gen_sample}.
The mesh generation process starts by overlaying a grid on the target area. It then produces a randomized tree covering
this grid. The individual mesh traces are then traced along a depth-first search through this tree. A visualization of
the steps is shown in Figure \ref{mesh_gen_viz}. A sample of the production results from our prototype is shown in
Figure \ref{mesh_gen_sample}.
\begin{figure}
\center
@ -413,14 +423,15 @@ shown in Figure \ref{mesh_gen_viz}. A sample of the production results from our
\subsection{Data transmission through rotating joint}
As a baseline solution for data transmission, we settled on a $\SI{115}{\kilo\baud}$ UART signal sent through a simple
bidirectional infrared link. In the transmitter, the UART TX line on-off modulates a $\SI{920}{\nano\meter}$ IR LED
through a common-emitter driver transistor. In the receiver, an IR PIN photodiode reverse-biased to
$\frac{1}{2}V_\text{CC}$ is connected to a reasonably wideband transimpedance amplifier (TIA) with a
$\SI{100}{\kilo\ohm}$ transimpedance. As shown in Figure \ref{photolink_schematic}, the output of this TIA is fed
through another $G=100$ amplifier whose output is then squared up by a comparator. We used an \texttt{MCP6494} quad
CMOS op-amp. At a specified $\SI{2}{\milli\ampere}$ current consumption it is within our rotor's power budget, and its
Gain Bandwidth Product of $\SI{7.5}{\mega\hertz}$ yields a useful transimpedance in the photodiode-facing TIA stage.
With the mesh done, the next engineering challenge was the mesh monitoring data link between rotor and stator. As a
baseline solution, we settled on a $\SI{115}{\kilo\baud}$ UART signal sent through a simple bidirectional infrared link.
In the transmitter, the UART TX line on-off modulates a $\SI{920}{\nano\meter}$ IR LED through a common-emitter driver
transistor. In the receiver, an IR PIN photodiode reverse-biased to $\frac{1}{2}V_\text{CC}$ is connected to a
reasonably wideband transimpedance amplifier (TIA) with a $\SI{100}{\kilo\ohm}$ transimpedance. As shown in Figure
\ref{photolink_schematic}, the output of this TIA is fed through another $G=100$ amplifier whose output is then squared
up by a comparator. We used an \texttt{MCP6494} quad CMOS op-amp. At a specified $\SI{2}{\milli\ampere}$ current
consumption it is within our rotor's power budget, and its Gain Bandwidth Product of $\SI{7.5}{\mega\hertz}$ yields a
useful transimpedance in the photodiode-facing TIA stage.
To reduce the requirements on power transmission to the rotor, we have tried to reduce power consumption of the
rotor-side receiver/transmitter pair trading off stator-side power consumption. One part of this is that we use
@ -446,8 +457,8 @@ driven at $\SI{1}{\milli\ampere}$ while the stator transmitter LED is driven at
\end{figure}
\subsection{Power transmission through rotating joint}
Since this prototype serves only demonstration purposes, we chose to use the simplest possible method of power
Besides the data link, the other electrical interface we need between rotor and stator is for power transmission. We
power Since this prototype serves only demonstration purposes, we chose to use the simplest possible method of power
transmission: solar cells. We mounted six series-connected solar cells in three commercially available modules on the
circular PCB at the end of our cylindrical rotor. The solar cells direclty feed the rotor's logic supply with buffering
by a large $\SI{33}{\micro\farad}$ ceramic capacitor. With six cells in series, they provide around $\SI{3.0}{\volt}$ at
@ -464,12 +475,15 @@ link.
\subsection{Evaluation}
During experiments, our prototype performed as intended. After some experimentation, we got both power and data
transmission through the rotating joint working reliably. Figure \ref{prototype_early_comms} shows our prototype
performing reliably at maximum speed for the first time. Our improvised IR link is open in both directions for about
$\SI{60}{\degree}$ of the rotation, which allows us to reliably transfer several tens of bytes in each direction during
each receiver's fly-by even at high speed of rotation. As a result of our prototype experiments, we consider a
larger-scale implementation of the inertial HSM concept practical.
After building our prototype inertial HSM according to the design decisions we outlined above, we performed a series of
experiments to validate the critical components of the design.
During these experiments, our prototype performed as intended. Both power and data transmission through the rotating
joint were working reliably. Figure \ref{prototype_early_comms} shows our prototype performing reliably at maximum speed
for the first time. Our improvised IR link is open in both directions for about $\SI{60}{\degree}$ of the rotation,
which allows us to reliably transfer several tens of bytes in each direction during the receivers' fly-by even at high
speed of rotation. As a result of our prototype experiments, we consider a larger-scale implementation of the inertial
HSM concept practical.
\begin{figure}
\center
@ -480,7 +494,9 @@ larger-scale implementation of the inertial HSM concept practical.
\label{prototype_early_comms}
\end{figure}
\section{Conclusion} \label{sec_conclusion} In this paper, we introduced inertial hardware security modules (iHSMs), a
\section{Conclusion}
\label{sec_conclusion} To conclude, in this paper we introduced inertial hardware security modules (iHSMs), a
novel concept for the construction of highly secure hardware security modules from inexpensive, commonly available
parts. We elaborated the engineering considerations underlying a practical implementation of this concept. We
implemented a prototype demonstrating practical solutions to the significant engineering challenges of this concept. We