ihsm/blogpost/ihsm-hiig-blog.txt

IHSMs for Grassroots Digital Autonomy
=====================================

In today's digital age, all of us depend on a number of large companies to mean well with us. To guard our data, or at
least not to abuse it too much without asking. Even if we wanted, we cannot take things into our own hands and reclaim
our digital autonomy because too much depends on too many system operators. To help small-time users and sysadmins to
fight this dependence, Jan Götte and Björn Scheuermann from HIIG have developed Inertial Hardware Security Modules, or
IHSMs. IHSMs enable low-budget servers to provide high security anywhere, even outside of well-protected datacenters.


Digital Dependencies
--------------------

The rise of networked computers has revolutionized many aspects of our modern lives. However, in lock-step with all the
quality of life improvements that the internet brought us, today we can observe a worrying trend of increasing digital
dependency. *Opting out* of the digital life on social media or e-commerce platforms becomes increasingly difficult to
sustain all the while the power of these platforms over their users is steadily increasing.  Anyone who has ever had
their Twitter or Facebook account suspended for an alleged Terms of Service violation or who has been locked out of
their Google or Amaazon account pending some complicated and frustrating “account verification” process for unnamed
“security reasons” will know how acute this dependency is. What the platform decides is law. There is no way for us as
users to meaningfully challenge its decisions because we are the platform's product, not its customers.

Don't it Yourself
-----------------

The obvious solution to this state of digital dependency is for us to take things into our own hands. On this premise,
open-source projects such as Nextcloud <https://nextcloud.com/> and Mastodon <https://joinmastodon.org/> have created
de-centralized alternatives to some of the big platforms. But DIY'ing has a drawback beyond the time spent on installing
and maintaining these solutions. While open-source software has made great strides over the last decades and offers
viable alternatives to many proprietary, centralized platforms today, this software has to run on someone's computer.
This underlying physical infrastructure is where things get complicated. Today, the only real option to run one's own
digital infrastructure is to rent a server (or part of one) from one of a number of large cloud providers. And in this
case, whoever runs the data center ultimately controls access to the data stored within.

At first, this fact of life might not seem concerning. After all, a data center operator would risk a lot if it
compromised its customers' security. However, in the past we have seen several physical attacks targeting high-value
data <https://archive.md/QpvHA> <https://en.wikipedia.org/wiki/NSA_ANT_catalog>. Large companies can mitigate such
attacks by tightly controlling their entire infrastructure, from the application software down to the data center's
access control systems. For smaller organizations and private individuals, this option does not exist.

Inertial Hardware Security Modules
----------------------------------

# Gaining back control over our digital lives is a monumental task. In software, through an ecosystem of open source
# projects, we have seen steady progress throughout the last decades. But all of that open source software needs to run
# somewhere, and today that usually means renting processing time at one of the large cloud providers. Avoiding them means
# buying your own hardware, and even then, that hardware needs to be put somewhere in a data center operated by yet
# another large company. Security-wise, this is not great since whoever runs the cloud or the datacenter can technically
# access everything running inside it. In the end it comes down to physical access: Even passwords or strong encryption
# cannot (yet) prevent someone with physical access to a machine from learning everything about its contents.

At HIIG, Björn and I have asked ourselves if there might be ways to claw back control over our hardware.  If we can find
a way for someone someone much smaller than an Amazon or Google to build a special physically secure server that is
invulnerable to physical attaks, this server can be put into any commercial data center while completely decoupling the
security of the data stored in it from the operator of the data center around it.

With Inertial Hardware Security Modules (short, IHSMs) we believe we have developed a strong approach towards a solution
to this problem. You can read the full paper here <https://tches.iacr.org/index.php/TCHES/article/view/9290>. The gist
of our solution is that we are able to build an enclosure that physically protects a server from any attempt at
siphoning off its secrets, no matter who tries to attack it. IHSMs are similar to a type of device used in payment
processing applications called Hardware Security Modules. Both provide a similar level of security. The main difference
is that IHSMs allow much larger systems to be protected and bump both size and computing power from that of a smartphone
to that of a server, a difference of about 100 × in space, processing power and cooling capacity.  IHSMs are simple to
construct and can be built with basic tools that can be found in most electronic labs. We envision a world where small
organizations can build their own secure servers that even a hostile secret service will not be able to compromise
physically.

IHSMs work by putting the server's mainboard, CPU, and memory inside of a tamper detection “mesh.” This mesh is a very
delicate component that will sense when someone tries to cut, drill or poke through it. When tampering is detected, an
alarm is activated and all sensitive data inside the server is deleted. To make the tamper detection mesh itself hard to
tamper with, this mesh is spinning at around the speed of a washing machine during its spin cycle.

Trusting and Un-trusting
------------------------

In computer security terms, what we are doing with IHSMs is that we are “untrusting” the server's physical environment.
Computer security has a funny definition of trust that is pretty much the opposite of everyone else. In computer
security, trust is a de-facto property of something that we de-facto rely on being good. In computer security, something
trusted should be trustworthy (but does not have to be). Contrast that to the common-sense meaning, where something
trustworthy should be trusted (but does not have to be).

Untrusting the data center gives us the chance to decouple software security from physical security. Using an IHSM, a
small organization such as the HIIG, the Chaos Computer Club, or a small company can operate software on a physical
foundation that is as secure as that of the large cloud companies, but with no one but themselves holding the keys to
the data.

Examples for applications where this increased security matters are group chat, cloud photo storage, and calendar and
contact synchronization systems. All of these already encrypt the network connection between their users phone or laptop
and the server, but only rarely encrypt the actual data stored on these servers.

# In terms of computer security, these platform behemoths are trusted parts of our lives. Not trusted as in
# “Google/Facebook/Amazon is a trustworthy company”, but trusted in the computer security sense of the world: they are
# de-facto trusted. We act as if we trust them, without actually considering whether they really are worth our trust. We
# need them, and we depend on them not acting against our interests. But if they abuse our trust, there is nothing we can
# do about it if we even notice.
# 
# Confusingly, this computer security meaning of trustedness is the exact opposite of our common-sense understanding of
# trust. In a common-sense world, a trustworthy thing should be trusted (but does not have to be). In the computer
# security world, a trusted thing should be trustworthy (but does not have to be). While the two seem close at first,
# there is a large difference in meaning. Understanding this difference is necessary to understand the things computer
# security people say.
# 
# The computer security concept of de-facto trust is counter-intuitve. However, its complement, which has no common-sense
# counterpart, is very useful. In computer security, the opposite of a trusted component is an untrusted component.
# Untrusted does not mean untrustworthy. An untrusted part is simply one that does not *need to be* trusted for things to
# work. An example for an untrusted part is a coffeeshop wifi network. While we need it to get online, we do not have to
# trust it to guard our data in any way since all our connections are authenticated and encrypted anyway using the TLS
# protocol that underlies HTTPS. In this case, by basing everything on an authenticated and encrypted network protocol, we
# remove trust from the coffeeshop wifi. We still need the wifi, but because our connections on top of it are encrypted
# separately, we do not have to care anymore whether someone is snooping on it or not.
# 
# From a computer security point of view, untrusting as many parts of a system as possible is highly desirable. The fewer
# parts you trust, the fewer you can concentrate your security engineering efforts on. Not only does this save time and
# money, the overall system will actually become more secure by doing this as more time is spent on less parts, leading to
# better-quality results.

# opposite of what everybody else calls trust. Take the sentence “Riley trusts Alex to take care of his cat.” In an
# everyday, common-sense way this sentence parses quite easily. Apparently Riley is out of town for a vacation or
# work-related trip and needs someone to feed their cat. Among Riley's friends, Riley asked Alex because they are
# reliable. The key concept here is that this sentence implies that *because Alex is trustworthy*, Riley asked them to
# take care of their cat.
# 
# In computer-land, trust means something very different. While in the common-sense interpretation trusting something
# implies that that thing is “trustworthy”, in computer security things are exactly backwards. In computer security, a
# trusted part of a system is one that has to be trusted because without it the system doesn't work. The distinction here
# is that this part *should* better be “trustworthy”, but that is not a given.
# 
# Now, why should we care about this rather pedantic distinction? That is because we have to keep in mind this linguistic
# quirk to correctly understand what security people mean when they are talking about computers. If we interpret things a
# computer security person says in a common-sense way, they might seem to make sense at first, but this important detail
# can get lost in translation.
# 
# There is a useful concept in computer security called “untrustedness”. In computer security, a thing is untrusted when
# we do not rely on it being trustworthy.


#The concept of a trusted thing as “something that must be trustworthy” is useful in computer security, but its
#complement is even more interesting. If there 


# The rise of networked computers has revolutionized many aspects of our modern lives. For most of us, the internet has
# made everyday things more convenient and improved our overall quality of life. Large platform companies can serve
# millions or even billions of users with only a few hundred people in design and development. However, these same
# economies of scale that have brought us incredibly polished software products that improve our lives in many ways, also
# have a dark side.
# 
# Through the last few decades, we can observe a shift in how we use computers in our lives. The computer systems we use
# have become increasingly complex and have proliferated throughout all aspects of our modern lives. At the same time, the
# internet's economies of scale have led to the operators of these systems consolidating. For instance, while in the 2000s
# an online business would have had to operate its own physical servers in a datacenter somewhere, today the entire
# internet rests on a small number of cloud computing providers.
# 
# On one axis, there is the shift towards increasing integration of digital systems into our lives. On the other axis
# there is the shift towards large centralized operators providing the infrastructure for most of these systems. These two
# effects compound each other to produce a steeply rising dependence towards system operators. On a national level, this
# is seen as a threat of an emerging concept dubbed “digital sovereignty”, and is perceived to be an acute issue. However,
# on an individual level, things are not better in any meaningful way--they are just talked about less.
#