In my most recent post, I described the first of two pending mandatory US cybersecurity regulations for IoT devices, the IoT Act of 2020. Now I’ll describe the second of these and the one that’s more important, the IoT “device labeling” requirement in the May 12 Executive Order on securing software sold to federal government agencies.
Paragraph (s) of section 4, on page
18 of the EO, reads: “The Secretary of Commerce, acting through the
Director of NIST, in coordination with representatives of other agencies as the
Director of NIST deems appropriate, shall initiate pilot programs informed by existing
consumer product labeling programs to educate the public on
the security capabilities of Internet-of-Things (IoT)
devices and software
development practices, and shall consider ways to incentivize
manufacturers and developers to participate in these programs.”
Paragraph (t) on page 19 reads: “Within
270 days of the date of this order, the Secretary of Commerce acting through
the Director of NIST, in coordination with the Chair of the Federal Trade
Commission (FTC) and representatives of other agencies as the Director of NIST
deems appropriate, shall identify IoT cybersecurity criteria for a consumer labeling
program, and shall consider whether such a consumer labeling program may be
operated in conjunction with or modeled after any similar existing government
programs consistent with applicable law. The criteria shall reflect
increasingly comprehensive levels of testing and assessment that a product may
have undergone and shall use or be compatible with existing labeling schemes
that manufacturers use to inform consumers about the security of their
products. The Director of NIST shall examine all relevant information,
labeling, and incentive programs and employ best practices. This review shall
focus on ease of use for consumers and a determination of what measures can be
taken to maximize manufacturer participation.”
Note that paragraph (u) mandates
a labeling program for consumer software. This is similar to the IoT device
labeling program and indeed, NIST is addressing these two programs together
(e.g. they’ve scheduled a two-day conference in September that will address
both programs).
What first struck me when I read these
two paragraphs was that putting labels on devices seems to be an odd way to
address cybersecurity. I normally think of a label as an indication that an
organization has tested a product and finds it meets certain safety standards. In
other words, the product is safe to use.
Cybersecurity (or security in
general) can’t be reduced to a simple “safe/unsafe” decision. In fact, let’s be
honest: Seldom (and perhaps never) in our personal or business lives do most of
us decide to buy a software product or an intelligent device, based solely on a
judgment of the product’s level of cybersecurity.
Instead, cybersecurity is at most
a final gating factor: We make the decision to buy a particular product, and
before we place the order (or put it in our shopping cart, whether physical or
virtual), we may ask ourselves whether there are any cybersecurity
skeletons in the closet of the supplier (a device manufacturer, a software
developer, the online community that developed an open source software product,
etc.) that are serious enough to warrant not going ahead with the purchase or
procurement. But recent history tells us that even a serious cybersecurity
breach that had a serious impact on customers (think Target or SolarWinds) in
the end has very little impact on demand for what the breached company sells –
and in fact the impact might be net positive if the company responded well to
the attack (as did both Target and SolarWinds).
So what we’re really looking for
regarding the security of an IoT device (or any other product containing or
consisting of software or firmware) is an indication of the risks that we are
going to undertake if we buy the product. If we know what those risks are, we
can then take steps to mitigate at least some of them and accept the rest. And if
we buy the product without even bothering to learn about the risks (which I’ll
admit is my approach when purchasing products for personal use. My general
reasoning is, “Given that well-known manufacturer XYZ develops this product
and/or that well-known retailer ABC sells the product, it must be secure enough
for my purposes.” Whether that’s good reasoning or not is left as an exercise
for the reader), we’re accepting those risks (although we all presumably take
some reasonable precautions with any product we buy, like at least set a
somewhat secure password, when we can do that).
So the main purpose of supply
chain cybersecurity in general (and not just for IoT devices) is to help us
understand the risks we undertook when we made the decision to purchase the product,
not to decide whether or not to purchase the product in the first place; we may
or may not take steps to mitigate any of those risks (or even to learn about
them), but if we want to do that, we can. And that’s why the idea of a device
label for cybersecurity seemed strange when I read about it in the EO: Since a
buy/no-buy decision very rarely depends on cybersecurity considerations, how
can a label help you determine what the specific risks are that apply to the
device, as well as what steps you could take to mitigate those risks? That’s
too much information to fit on a label.
It turns out, although I didn’t
know this until recently, that the idea of device labeling for IoT device
cybersecurity has been around for a while in Europe. In fact, perhaps the best
such program (albeit one that’s almost infinitesimally small compared with the program
that will be required in the US) is one introduced by the Finnish government in
2019 (and no bad puns about whether or not the program is Finnished or whether
or not it only applies to Finnished products. Bad puns are forbidden in this blog.
I can assure you that I’ll never sink to using this low form of humor, such as asking
whether the army fights to the (last) Finnish. After all, what kind of blogger do
you think I am?).
There are three main components to
the Finnish IoT device labeling program:
First, the manufacturer has to certify
that the device complies with 18 of the 70-odd requirements, which are referred
to as “Provisions”, in the ETSI
303 645 standard for IoT devices. This standard may become mandatory in
Europe at some point (in the sense that the European Parliament will enact
legislation requiring it). But more importantly, this will almost certainly
become a de facto standard for commercial, industrial and home-based IoT
devices in Europe.
It may seem odd that the
manufacturer has to certify compliance with just 18 of the requirements of ETSI
303 645. Why not all of them? For one thing, since Finland is part of the
European Union, their IoT manufacturers will ultimately have to be in
compliance with all of the standard anyway. But the real reason is that whoever
drafted this law understood that the term “IoT device” covers a huge range of products,
from baby monitors and doorbell cameras to protective relays and RTUs (remote
terminal units).
This means that, for any
particular IoT device, a large portion of the provisions in ETSI 303 645 won’t
apply. The 18 requirements chosen by the Finnish were sufficiently general,
that it can reasonably be expected that they’ll all apply to the great majority
of IoT products. Examples of the 18 provisions are “a manufacturer
should have a ‘public vulnerability disclosure policy’, ‘installation of
updates should be secure’, ‘all unused interfaces must be disabled’, and ‘there
should be no hard coded credentials in the system’.”
You might well ask, “How much good
does it do for the user to know that the device they bought meets just 18 very
general – and fairly easy to comply with – requirements?” My answer is “Probably
not much.” However, the second step in the program directly addresses this gap.
Instead of simply requiring that the organization comply with all of the ETIS
303 645 provisions or something like that, it requires that a third-party “inspecting
body”, which is approved by a government agency called Traficom, should do a
threat-modeling exercise for the device.
The term “threat modeling” is one
of those terms that’s thrown around a lot and is used with a variety of
meanings – so it’s almost meaningless (no pun intended here. Remember: This
blog doesn’t do puns!) to discuss the meaning of threat modeling in general. However,
in the context of the Finnish device labeling requirement, the term means that the
inspecting body needs to consider both what the device will be used for (e.g. a
baby monitor isn’t usually used for purposes that require high security, but a
protective relay usually is. On the other hand, protecting personal information
is a big concern for baby monitors, but it’s an almost nonexistent concern –
except for login information, of course – for protective relays), as well as
the setting in which it will be used (e.g. a room in a private home in the case
of a baby monitor, vs. a Transmission substation where a cyberattack might affect
thousands or even millions of people, in the case of the relay).
Using these two considerations,
the inspecting body needs to identify all important cybersecurity threats that might
apply to that device. Some of them will apply to many devices (like password strength),
while others may be specific to the particular device. For example, one significant
threat that applies to smart speaker devices like Amazon Echo™ is that, because
they are continually listening to conversations for the magic word “Alexa”, if
compromised they might be used to eavesdrop on conversations that occur in the
home. In fact, DoD recently put out an advisory to employees working from home
not to have smart speakers in the same room where they are working.
Third, when the inspecting body
has identified the important cyber threats that apply to the product, they need
to develop a test plan to determine whether each threat poses an actual risk.
In my way of speaking, if a threat poses a risk, this means either the
likelihood or impact of the threat being realized is high.
Since the device can have many
uses, it’s very hard – nay, in my opinion it’s impossible – to determine the
degree of impact of a threat being realized in all cases. Therefore, what’s
important is likelihood, which can be high or low. If likelihood is high, the
risk of the threat being realized is high, and mitigation measures should be taken
to reduce the likelihood to low. And if likelihood is low, this means the risk
is already mitigated (which might be by Mother Nature or other completely
external forces, e.g. the risk that a forest fire will damage a region of the
Sahara Desert is about as close to zero as it could be), and no further
mitigation measures are warranted.
For example, in the case of the Amazon
Echo threat discussed earlier, the inspecting body might decide that this threat
is low, due to controls that Amazon has implemented to protect the audio stream
the device “hears”. In this case, the body will deem this threat to pose
minimal risk, meaning users need not take any further mitigation measures for
it. So this threat will be deemed not to pose a risk in the Echo device, and it
won’t be included in the test plan.
In the Finnish program, when the
inspecting body has developed their test plan, they must submit it to Traficom
(the government agency that operates the device labeling program). Traficom can
disapprove the plan, in which case the inspecting body needs to revise it and
resubmit it; otherwise, the plan is approved, and the inspecting body can go
ahead and conduct the testing.
Finally, the testing body conducts
the tests and submits the test plan and the results to Traficom. Traficom reviews
these, then decides whether or not to grant the label. Traficom decides whether
whatever residual risk remains (after the results of the device testing are
considered) allows the device to be sold in Finland.
I think the Finnish model would be
an excellent one for the US program, with two important modifications:
First, the US program has to
operate with minimal government involvement, both for budgetary reasons and for
general regulatory philosophical reasons (as long as human life isn’t directly at
stake, the feds usually don’t want to be directly involved in auditing or
testing. So the Consumer Product Safety Commission and the National Highway
Traffic Safety Administration are directly involved in investigating cases
that come up for them, since these could often have an impact on life safety. But
it can’t be said that the cyber compromise of an IoT device would directly
impact human life, except in rare cases (e.g. if a baby died because his or her
parents couldn’t hear their distress cries, due to a cyberattack having
disrupted communications with the baby monitor).
This means that the Office of Management
and Budget, which is charged with implementing all of the EO’s provisions[i], won’t want to be in the
business of certifying “inspecting bodies” – which may be called “laboratories”,
as they are in Europe. Rather, they would certify the organizations that certify
the labs, and might even go one step further: certify the organizations that
certify the organizations that certify the labs.[ii]
The second difference between the
Finnish IoT device labeling program and the likely US program is that the latter
won’t be required for a device to be sold in the US, or even to the federal
government. A manufacturer will be able to decide for themselves whether or not
they want to participate in the program. If they decide not to, they won’t be
shut out of the market, but they will have a harder time selling their product –
especially to federal agencies, since the agencies will be required to ask
their IoT device suppliers to participate in the labeling program, and they’ll
have to justify to auditors any case in which they bought from a supplier that
didn’t participate (i.e. the supplier doesn’t have a label at all, regardless
of what it says).
So what will the “label” say, if it isn’t “This device is safe/unsafe to buy”? The label (which will probably be a document accessible by scanning a QR code or entering a URL found on the product or in its accompanying documentation) will list the threats that were identified through threat modeling and the degree of residual risk for each threat, after any mitigations that have been applied by the manufacturer. Additionally, the document will list mitigations that a user organization, which is concerned about one of the risks, can take to at least partially mitigate it. For example, if the organization (a federal agency, a private organization, etc.) wishes to partially mitigate the risk due to the fact that a device allows users to authenticate using weak passwords, they could locate the devices only in access-controlled rooms.
Any opinions expressed in this blog post are strictly mine and are not necessarily shared by any of the clients of Tom Alrich LLC. Nor are they shared by the National Technology and Information Administration’s Software Component Transparency Initiative, for which I volunteer as co-leader of the Energy SBOM Proof of Concept. If you would like to comment on what you have read here, I would love to hear from you. Please email me at tom@tomalrich.com.
[i] Although the EO allows OMB to decide whether to bring in another agency to run any particular program. You’ll notice that paragraph 4 (t) quoted above mentions the Federal Trade Commission (FTC) as being involved with the device labeling program. This makes a lot of sense, and comports with my idea that, sooner or later, all government agencies will be involved with cybersecurity, and not just for securing their own systems.
[ii] This
is essentially how the EPA’s Energy
Star program works: the EPA approves Certification Bodies. The CBs then
approve Accreditation Bodies, who accredit the laboratories that test products
for energy consumption. The Accreditation Bodies accredit based on ISO/IEC 17025, an international standard for testing
laboratories (which is independent of the subject being tested). This is the basis
for cybersecurity laboratory testing in Europe, and most likely will be the basis here as
well, for the IoT device labeling program.
No comments:
Post a Comment