Defining the Edge

How might we differentiate Edge computing from Cloud computing?

This is an edited excerpt from my forthcoming book on Trust in Computing and the Cloud for Wiley.

There’s been a lot of talk about the Edge, and almost as many definitions as there are articles out there. As usual on this blog, my main interest is around trust and security, so this brief look at the Edge concentrates on those aspects, and particularly on how we might differentiate Edge computing from Cloud computing.

The first difference we might identify is that Edge computing addresses use cases where consolidating compute resource in a centralised location (the typical Cloud computing case) is not necessarily appropriate, and pushes some or all of the computing power out to the edges of the network, where it can computing resources can process data which is generated at the fringes, rather than having to transfer all the data over what may be low-bandwidth networks for processing. There is no generally accepted single industry definition of Edge computing, but examples might include:

  • placing video processing systems in or near a sports stadium for pre-processing to reduce the amount of raw footage that needs to be transmitted to a centralised data centre or studio
  • providing analysis and safety control systems on an ocean-based oil rig to reduce reliance and contention on an unreliable and potentially low-bandwidth network connection
  • creating an Internet of Things (IoT) gateway to process and analyse data environmental sensor units (IoT devices)
  • mobile edge computing, or multi-access edge computing (both abbreviated to MEC), where telecommunications services such as location and augmented reality (AR) applications are run on cellular base stations, rather than in the telecommunication provider’s centralised network location.

Unlike Cloud computing, where the hosting model is generally that computing resources are consumed by tenants – customers of a public cloud, for instance- in the Edge case, the consumer of computing resources is often the owner of the systems providing them (though this is not always the case). Another difference is the size of the host providing the computing resources, which may range from very large to very small (in the case of an IoT gateway, for example). One important factor about most modern Edge computing environments is that they employ the same virtualisation and orchestration techniques as the cloud, allowing more flexibility in deployment and lifecycle management over bare-metal deployments.

A table comparing the various properties typically associated with Cloud and Edge computing shows us a number of differences.


Public cloud computingPrivate cloud computingEdge computing
LocationCentralisedCentralisedDistributed
Hosting modelTenantsOwnerOwner or tenant(s)
Application typeGeneralisedGeneralisedMay be specialised
Host system sizeLargeLargeLarge to very small
Network bandwidthHighHighMedium to low
Network availabilityHighHighHigh to low
Host physical securityHighHighLow
Differences between Edge computing and public/private cloud computing

In the table, I’ve described two different types of cloud computing: public and private. The latter is sometimes characterised as on premises or on-prem computing, but the point here is that rather than deploying applications to dedicated hosts, workloads are deployed using the same virtualisation and orchestration techniques employed in the public cloud, the key difference being that the hosts and software are owned and managed by the owner of the applications. Sometimes these services are actually managed by an external party, but in this case there is a close commercial (and concomitant trust) relationship to this managed services provider, and, equally important, single tenancy is assured (assuming that security is maintained), as only applications from the owner of the service are hosted[1]. Many organisations will mix and match workloads to different cloud deployments, employing both public and private clouds (a deployment model known as hybrid cloud) and/or different public clouds (a deployment model known as multi-cloud). All these models – public computing, private computing and Edge computing – share an approach in common: in most cases, workloads are not deployed to bare-metal servers, but to virtualisation platforms.

Deployment model differences

What is special about each of the models and their offerings if we are looking at trust and security?

One characteristic that the three approaches share is scale: they all assume that hosts will have with multiple workloads per host – though the number of hosts and the actual size of the host systems is likely to be highest in the public cloud case, and lowest in the Edge case. It is this high workload density that makes public cloud computing in particular economically viable, and one of the reasons that it makes sense for organisations to deploy at least some of their workloads to public clouds, as Cloud Service Providers can employ economies of scale which allow them to schedule workloads onto their servers from multiple tenants, balancing load and bringing sets of servers in and out of commission (a computation- and time-costly exercise) infrequently. Owners and operators of private clouds, in contrast, need to ensure that they have sufficient resources available for possible maximum load at all times, and do not have the opportunities to balance loads from other tenants unless they open up their on premises deployment to other organisations, transforming themselves into Cloud Service Providers and putting them into direct competition with existing CSPs.

It is this push for high workload density which is one of the reasons for the need for strong workload-from-workload (type 1) isolation, as in order to be able to maintain high density, cloud owners need to be able to mix workloads from multiple tenants on the same host. Tenants are mutual untrusting; they are in fact likely to be completely unaware of each other, and, if the host is doing its job well, unaware of the presence of other workloads on the same host as them. More important than this property, however, is a strong assurance that their workloads will not be negatively impacted by workloads from other tenants. Although negative impact can occur in other contexts to computation – such as storage contention or network congestion – the focus is mainly on the isolation that hosts can provide.

The likelihood of malicious workloads increases with the number of tenants, but reduces significantly when the tenant is the same as the host owner – the case for private cloud deployments and some Edge deployments. Thus, the need for host-from-workload (type 2) isolation is higher for the public cloud – though the possibility of poorly written or compromised workloads means that it should not be neglected for the other types of deployment.

One final difference between the models is that for both public and private cloud deployments the physical vulnerability of hosts is generally considered to be low[2], whereas the opportunities for unauthorised physical access to Edge computing hosts are considered to be much higher. You can read a little more about the importance of hardware as part of the Trusted Compute Base in my article Turtles – and chains of trust, and it is a fundamental principle of computer security that if an attacker has physical access to a system, then the system must be considered compromised, as it is, in almost all cases, possible to compromise the confidentiality, integrity and availability of workloads executing on it.

All of the above are good reasons to apply Confidential Computing techniques not only to cloud computing, but to Edge computing as well: that’s a topic for another article.


1 – this is something of a simplification, but is a useful generalisation.

2 – Though this assumes that people with authorised access to physical machines are not malicious, a proposition which cannot be guaranteed, but for which monitoring can at least be put in place.

Arm joins the Confidential Computing party

Arm’s announcement of Realms isn’t just about the Edge

The Confidential Computing Consortium is a Linux Project designed to encourage open source projects around confidential computing. Arm has been part of the consortium for a while – in fact, the company is Premier Member – but things got interesting on the 30th March, 2021. That’s when Arm announced their latest architecture: Arm 9. Arm 9 includes a new set of features, called Realms. There’s not a huge amount of information in the announcement about Realms, but Arm is clear that this is their big play into Confidential Computing:

To address the greatest technology challenge today – securing the world’s data – the Armv9 roadmap introduces the Arm Confidential Compute Architecture (CCA).

I happen to live about 30 minutes’ drive from the main Arm campus in Cambridge (UK, of course), and know a number of Arm folks professionally and socially – I think I may even have interviewed for a job with them many moons ago – but I don’t want to write a puff piece about the company or the technology[1]. What I’m interested in, instead, is the impact this announcement is likely to have on the Confidential Computing landscape.

Arm has had an element in their architecture for a while called TrustZone which provides a number of capabilities around security, but TrustZone isn’t a TEE (Trusted Execution Environment) on its own. A TEE is the generally accepted unit of confidential computing – the minimum building block on which you can build. It is arguably possible to construct TEEs using TrustZone, but that’s not what it’s designed for, and Arm’s decision to introduce Realms strongly suggests that they want to address this. This is borne out by the press release.

Why is all this important? I suspect that few of you have laptops or desktops that run on Arm (Raspberry Pi machines apart – see below). Few of the servers in the public cloud run Arm, and Realms are probably not aimed particularly at your mobile phone (for which TrustZone is a better fit). Why, then, is Arm bothering to make a fuss about this and to put such an enormous design effort into this new technology? There are two answers, it seems to me, one of which is probably pretty much a sure thing, and the other of which is more of a competitive gamble.

Answer 1 – the Edge

Despite recent intrusions by both AMD and Intel into the Edge space, the market is dominated by Arm-based[3] devices. And Edge security is huge, partly because we’re just seeing a large increase in the number of Edge devices, and partly because security is really hard at the Edge, where devices are more difficult to defend, both logically (they’re on remote networks, more vulnerable to malicious attack) and physically (many are out of the control of their owners, living on customer premises, up utility poles, on gas pipelines or in sports stadia, just to give a few examples). One of the problems that confidential computing aims to solve is the issue that, traditionally, once an attacker has physical access to a system, it should be considered compromised. TEEs allow some strong mitigations against that problem (at least against most attackers and timeframes), so making it easy to create and use TEEs on the Edge makes a lot of sense. With the addition of Realms to the Arm 9 architecture, Arm is signally its intent to address security on the Edge, and to defend and consolidate its position as leader in the market.

Answer 2 – the Cloud

I mentioned above that few public cloud hosts run Arm – this is true, but it’s likely to change. Arm would certainly like to see it change, and to see its chipsets move into the cloud mainstream. There has been a lot of work to improve support for server-scale Arm within Linux (in fact, open source support for Arm is generally excellent, not least because of the success of Arm-based chips in Raspberry Pi machines). Amazon Cloud Services (AWS) started offering Arm-based servers to customers as long ago as 2018. This is a market in which Arm would clearly love to be more active and carve out a larger share, and the growing importance of confidential computing in the cloud (and public and private) means that having a strong story in this space was important: Realms are Arm’s answer to this.

What next?

An announcement of an architecture is not the same as availability of hardware or software to run on it. We can expect it to be quite a few months before we see production chips running Arm 9, though evaluation hardware should be available to trusted partners well before that, and software emulation for various components of the architecture will probably come even sooner. This means that those interested in working with Realms should be able to get things moving and have something ready pretty much by the time of availability of production hardware. We’ll need to see how easy they are to use, what performance impact they have, etc., but Arm do have an advantage here: as they are not the first into the confidential computing space, they’ve had the opportunity to watch Intel and AMD and see what has worked, and what hasn’t, both technically and in terms of what the market seems to like. I have high hopes for Arm Realms, and Enarx, the open source confidential computing project with which I’m closely involved, has plans to support them when we can: our architecture was designed with multi-platform support from the beginning.


1 – I should also note that I participated in a panel session on Confidential Computing which was put together by Arm for their “Arm Vision Day”, but I was in no way compensated for this[2].

2 -in fact, the still for the video is such a terrible picture of me that I think maybe I have grounds to sue for it to be taken down.

3 – Arm doesn’t manufacture chips itself: it licenses its designs to other companies, who create, manufacture and ship devices themselves.

Measured and trusted boot

What they give you – and don’t.

Sometimes I’m looking around for a subject to write about, and realise that there’s one which I assume that I’ve covered, but, on searching, discover that I haven’t. Such a one is “measured boot” and “trusted boot” – sometimes, misleadingly, referred to as “secure boot”. There are specific procedures which use these terms with capital letters – e.g. Secure Boot – which I’m going to try to avoid discussing in this post. I’m more interested in the generic processes, and a major potential downfall, than in trying to go into the ins and outs of specifics. What follows is a (heavily edited) excerpt from my forthcoming book on Trust in Computing and the Cloud for Wiley.

In order to understand what measured boot and trusted boot aim to achieve, let’s have a look at the Linux virtualisation stack: the components you run if you want to be using virtual machines (VMs) on a Linux machine. This description is arguably over-simplified, but we’re not interested here in the specifics (as I noted above), but in what we’re trying to achieve. We’ll concentrate on the bottom four layers (at a rather simple level of abstraction): CPU/management engine; BIOS/EFI; Firmware; and Hypervisor, but we’ll also consider a layer just above the CPU/management engine, where we interpose a TPM (a Trusted Platform Module) and some instructions for how to perform one of our two processes. Once the system starts to boot, the TPM is triggered, and then starts its work (alternative roots of trust such as HSMs might also be used, but we will use TPMs, the most common example in this context, as our example).

In both cases, the basic flow starts with the TPM performing a measurement of the BIOS/EFI layer. This measurement involves checking the binary instructions to be carried out by this layer, and then creating a cryptographic hash of the binary image. The hash that’s produced is then stored in one of several “PCR slots” in the TPM. These can be thought of as pieces of memory which can be read later on, either by the TPM for its purposes, or by entities external to the TPM, but which cannot be changed once they have been written. This provides assurances that once a value is written to a PCR by the TPM, it can be considered constant for the lifetime of the system until power-off or reboot.

After measuring the BIOS/EFI layer, the next layer (Firmware) is measured. In this case, the resulting hash is combined with the previous hash (which was stored in the PCR slot) and then itself stored in a PCR slot. The process continues until all of the layers involved in the process have been measured, and the results of the hashes stored. There are (sometimes quite complex) processes to set up the original TPM values (I’ve missed out some of the more low-level steps in the process for simplicity) and to allow (hopefully authorised) changes to the layers for upgrading or security patching, for example. What this process “measured boot” allows is for entities to query the TPM after the process has completed, and check whether the values in the PCR slots correspond to the expected values, pre-calculated with “known good” versions of the various layers – that is, pre-checked versions whose provenance and integrity have already been established. Various protocols exist to allow parties external to the system to check the values (e.g. via a network connection) that the TPM attests to being correct: the process of receiving and checking such values from an external system is known as “remote attestation”.

This process – measured boot – allows us to find out whether the underpinnings of our system – the lowest layers – are what we think they are, but what if they’re not? Measured boot (unsurprisingly, given the name) only measures, but doesn’t perform any other actions. The alternative, “trusted boot” goes a step further. When a trusted boot process is performed, the process not only measures each value, but also performs a check against a known (and expected!) good value at the same time. If the check fails, then the process will halt, and the booting of the system will fail. This may sound like a rather extreme approach to take to a system, but sometimes it is absolutely the right one. Where the system under consideration may have been compromised – which is one likely inference that you can make from the failure of a trusted boot process – then it is better that it not be available at all than to be running based on flawed expectations.

This is all very well if I’m the owner of the system which is being measured, have checked all of the various components being measured (and the measurements), and so can be happy that what’s being booted it what I want[1]. But what if I’m actually using a system on the cloud, for instance, or any system owned and managed by someone elese? In that case, I’m trusting the cloud provider (or owner/manager) with two things:

  1. do all the measuring correctly, and report correct results to me;
  2. actually to have built something which I should be trusting in the first place!

This is the problem with the nomenclature “trusted boot”, and, even worse, “secure boot”. Both imply that an absolute, objective property of a system has been established – it is “trusted” or “secure” – when this is clearly not the case. Obviously, it would be unfair to expect the designers of such processes to name them after the failure states – “untrusted boot” or “insecure boot” – but unless I can be very certain that I trust the owner of the system to do step 2 entirely correctly (and in my best interests, as user of the system, rather than theirs, and owner) then we can make no stronger assertions. There is an enormous temptation to take a system which has gone through a trusted boot process and to label it a “trusted system”, where the very best assertion we can make is that the particular layers measured in the measured and/or trusted boot process have been asserted to be those which the process expected to be present. Such a process says nothing at all about the fitness of the layers to provide assurances of behaviour, nor about the correctness (or fitness to provide assurances of behaviour) of any subsequent layers on top of those.

It’s important to note that designers of TPMs are quite clear what is being asserted, and that assertions about trust should be made carefully and sparingly. Unluckily, however, the complexities of systems, the general low level of understanding of trust, and the complexities of context and transitive trust make it very easy for designers and implementors of systems to do the wrong thing, and to assume that any system which has successfully performed a trusted boot process can be considered “trusted”. It is also extremely important to remember that TPMs, as hardware roots of trust, offer us one of the best mechanisms for we have for establishing a chain of trust in systems that we may be designing or implementing, and I plan to write an article about them soon.


1 – although this turns out to be much harder to do that you might expect!

Vint Cerf’s “game changer”

I’m really proud to be involved with a movement which I believe can change the way we do computing.

Today’s article is a little self-indulgent, but please bear with me, as I’m a little excited. Vint Cerf is one of a small handful of people who have a claim to being called “greats”. He’s one of the co-developers of TCP/IP protocol with Bob Kahn in 1974, and has been working on technology – much of it pretty cool technology – since then. I turned 50 recently, and if I’d achieved half of what he had by his 50th birthday, I’d be feeling more accomplished than I do right now! As well as his work in technology, he’s also an advocate for accessibility, which is something which is also dear to my heart.

What does this have to do with Alice, Eve and Bob – a security blog? Well, last week, Dark Reading[1], an influential technology security site, published a commentary piece by Cerf under its “Cloud” heading: Why Confidential Computing is a Game Changer. I could hardly have been more pleased: this is an area which I’m very excited about, and which the Enarx project, of which I’m co-founder, addresses. The Enarx project is part of the Confidential Computing Consortium (mentioned in Cerf’s article), a Linux Foundation project to increase use of confidential computing through open source projects.

So, what is confidential computing? Cerf describes it as “a breakthrough technology that encrypts data in use, while it is being processed”. He goes on to give a good description of the technology, noting that Google (his employer[2]) has recently released a product using confidential computing. Google is actually far from the first cloud service provider to do this, but it’s only fair that Cerf should mention his employer’s services from time to time: I’m going to forgive him, given how enthusiastic he is about the technology more generally. He describes it as a transformational technology which “will and should be a part of every enterprise cloud deployment”.

I agree, and it’s really exciting to see such a luminary embracing the possibilities the confidential computing presents. For those readers who aren’t aware of what it is, confidential computing allows you to keep data and processes secret in the cloud, on private servers, on the Edge, IoT, etc. – even from administrators, hypervisors and the host kernel. It uses TEEs – Trusted Execution Environments – to protect the confidentiality and integrity of the workloads (application, programs) that you want to run. If you’re not sure you trust your cloud provider, if your regulatory body won’t let you run your applications in certain places, if you want to deploy to machines which are vulnerable to attack – physical or logical – then TEEs and confidential computing can help.

You can find a more information in some of my articles:

You can always visit the Confidential Computing Consortium[3] or visit the Enarx project (links above): all of our code and documentation is open, and we’d love to see you. I’m really proud to be involved with – in fact, deeply embedded in – a movement which I believe can change the way we do computing. And really excited that someone like Vint Cerf agrees.


1 – I have no affiliation with Dark Reading, though I do recommend it to readers of this blog.

2- neither do I have any affiliation with Google or Alphabet, its parent!

3 – I am, however, a member of both the Governing Board and the Technical Advisory Council of the Confidential Computing Consortium. I’m also the Treasurer.

Isolationism – not a 4 letter word (in the cloud)

Things are looking up if you’re interested in protecting your workloads.

In the world of international relations, economics and fiscal policy, isolationism doesn’t have a great reputation. I could go on, I suppose, if I did some research, but this is a security blog[1], and international relations, fascinating area of study though it is, isn’t my area of expertise: what I’d like to do is borrow the word and apply it to a different field: computing, and specifically cloud computing.

In computing, isolation is a set of techniques to protect a process, application or component from another (or a set of the former from a set of the latter). This is pretty much always a good thing – you don’t want another process interfering with the correct workings of your one, whether that’s by design (it’s malicious) or in error (because it’s badly designed or implemented). Isolationism, therefore, however unpopular it may be on the world stage, is a policy that you generally want to adopt for your applications, wherever they’re running.

This is particularly important in the “cloud”. Cloud computing is where you run your applications or processes on shared infrastructure. If you own that infrastructure, then you might call that a “private cloud”, and infrastructure owned by other people a “public cloud”, but when people say “cloud” on its own, they generally mean public clouds, such as those operated by Amazon, Microsoft, IBM, Alibaba or others.

There’s a useful adage around cloud computing: “Remember that the cloud is just somebody else’s computer”. In other words, it’s still just hardware and software running somewhere, it’s just not being run by you. Another important thing to remember about cloud computing is that when you run your applications – let’s call them “workloads” from here on in – on somebody else’s cloud (computer), they’re unlikely to be running on their own. They’re likely to be running on the same physical hardware as workloads from other users (or “tenants”) of that provider’s services. These two realisations – that your workload is on somebody else’s computer, and that it’s sharing that computer with workloads from other people – is where isolation comes into the picture.

Workload from workload isolation

Let’s start with the sharing problem. You want to ensure that your workloads run as you expect them to do, which means that you don’t want other workloads impacting on how yours run. You want them to be protected from interference, and that’s where isolation comes in. A workload running in a Linux container or a Virtual Machine (VM) is isolated from other workloads by hardware and/or software controls, which try to ensure (generally very successfully!) that your workload receives the amount of computing time it should have, that it can send and receive network packets, write to storage and the rest without interruption from another workload. Equally important, the confidentiality and integrity of its resources should be protected, so that another workload can’t look into its memory and/or change it.

The means to do this are well known and fairly mature, and the building blocks of containers and VMs, for instance, are augmented by software like KVM or Xen (both open source hypervisors) or like SELinux (an open source capabilities management framework). The cloud service providers are definitely keen to ensure that you get a fair allocation of resources and that they are protected from the workloads of other tenants, so providing workload from workload isolation is in their best interests.

Host from workload isolation

Next is isolating the host from the workload. Cloud service providers absolutely do not want workloads “breaking out” of their isolation and doing bad things – again, whether by accident or design. If one of a cloud service provider’s host machines is compromised by a workload, not only can that workload possibly impact other workloads on that host, but also the host itself, other hosts and the more general infrastructure that allows the cloud service provider to run workloads for their tenants and, in the final analysis, make money.

Luckily, again, there are well-known and mature ways to provide host from workload isolation using many of the same tools noted above. As with workload from workload isolation, cloud service providers absolutely do not want their own infrastructure compromised, so they are, of course, going to make sure that this is well implemented.

Workload from host isolation

Workload from host isolation is more tricky. A lot more tricky. This is protecting your workload from the cloud service provider, who controls the computer – the host – on which your workload is running. The way that workloads run – execute – is such that such isolation is almost impossible with standard techniques (containers, VMs, etc.) on their own, so providing ways to ensure and prove that the cloud service provider – or their sysadmins, or any compromised hosts on their network – cannot interfere with your workload is difficult.

You might expect me to say that providing this sort of isolation is something that cloud service providers don’t care about, as they feel that their tenants should trust them to run their workloads and just get on with it. Until sometime last year, that might have been my view, but it turns out to be wrong. Cloud service providers care about protecting your workloads from the host because it allows them to make more money. Currently, there are lots of workloads which are considered too sensitive to be run on public clouds – think financial, health, government, legal, … – often due to industry regulation. If cloud service providers could provide sufficient isolation of workloads from the host to convince tenants – and industry regulators – that such workloads can be safely run in the public cloud, then they get more business. And they can probably charge more for these protections as well! That doesn’t mean that isolating your workloads from their hosts is easy, though.

There is good news, however, for both cloud service providers and their teants, which is that there’s a new set of hardware techniques called TEEs – Trusted Execution Environments – which can provide exactly this sort of protection[2]. This is rapidly maturing technology, and TEEs are not easy to use – in that it can not only be difficult to run your workload in a TEE, but also to ensure that it’s running in a TEE – but when done right, they do provide the sorts of isolation from the host that a workload wants in order to maintain its integrity and confidentiality[3].

There are a number of projects looking to make using TEEs easier – I’d point to Enarx in particular – and even an industry consortium to promote open TEE adoption, the Confidential Computing Consortium. Things are looking up if you’re interested in protecting your workloads, and the cloud service providers are on board, too.


1 – sorry if you came here expecting something different, but do stick around and have a read: hopefully there’s something of interest.

2 – the best known are Intel’s SGX and AMD’s SEV.

3 – availability – ensuring that it runs fairly – is more difficult, but as this is a property that is also generally in the cloud service provider’s best interest, and something that can can control, it’s not generally too much of a concern[4].

4 – yes, there are definitely times when it is, but that’s a story for another article.