Internet of Things: Who and what can you trust?

7 mins read

It depends upon who you talk to and their interpretation of what the Internet of Things actually is, but in the next decade or so, there could be trillions of devices with the potential to communicate with each other.

With this degree of complexity, it's no surprise to find three words are being used with increasing frequency – trust, security and authentication. Trust relates to whether the data being sent is what you expect. Security relates to making sure that your data doesn't go where you don't expect it to. Authentication means finding out whether the device being talked with is actually the one that is expected. So what is the IoT? How does it differ, if at all, from M2M? Paul Green is director of innovation for Arkessa, an M2M and IoT specialist recently recognised by Gartner as a 'cool vendor in the IoT'. "Most people get into trouble when they try to work out which one is which," he said. "Many M2M companies changed to using IoT because they thought it sounded better. "Something like a supermarket chain or a water company is in the M2M world because they are closed systems," he explained. "Data is being taken from 'things' and into their database. The applications used are essentially silo based; lots of things feed into one engine. "The IoT is entirely different," he contended. "Think about something like a toaster – who could possibly want data from that? Well, it could be the manufacturer, looking at failure rates; the brand owner; it could be someone selling bread. It could also be a health visitor or a family member, interested in knowing the user is 'up and about'. It could be the electricity company. "A toaster is just a dumb device that can only tell you if it's on or off," he continued, "but it's something which can be used by a range of people." Iain Davidson, European networks marketing manager for Freescale, noted: "I tend to think of M2M and IoT as being the same, but differentiated by who deploys and who uses the data. M2M is a closed system, but IoT is more powerful; once it's connected, anyone can subscribe." Charlene Marini is vice president of marketing, embedded segments, for ARM; a company which has the IoT firmly in its sights. Last year, it turned its Cambridge HQ into an 'IoT hub' and bought Finland based Sensinode, an IoT software specialist. She said: "M2M is a more proprietary deployment, intended for one type of service. IoT, by contrast, is built on web enabled standards that offer interoperability, scalability and services. Both have different requirements and there is an evolution underway as M2M moves to IoT, but some industries will stay with M2M." The difference between M2M and IoT lies not in the data, but in the way in which the data is made available and used. "Sources of data are being used by lots of different people," Green continued, "but they are seeing the data in different ways; the difference comes in the way in which the data is interpreted." Arkessa has grown from developing remote communication systems for vending machines to its current focus on making M2M and IoT systems easy to implement and integrate. "Each of our original customers needed communications to work across Europe," Green recalled. "Making those systems work was difficult, especially in 2000." Arkessa's M2M business started in 2009. "All the street lighting in Central London is controlled over our network," Green said, "as are many motorway information signs. But that's M2M, not IoT. With IoT, instead of things connecting only to your 'back office', there are devices out there saying 'I'm a thing; what do you want to do?'." While many of the issues relating to the IoT are to do with hardware, many can be seen as 'soft'. Marini said: "The IoT will need multilayer solutions. Implementation will be a significant portion; how you can implement secure memory? Then, once a device is in the system, how does it communicate with the real world? "Standards will be needed at different levels and one of the biggest issues will be simplicity, because there will be multiple layers in an IoT implementation. Industry needs to look at how to simplify this and how to deliver security and trust." Davidson noted that Freescale has a security centre of excellence in Texas. "The technology being developed there is finding its way into MCU type products." Arkessa is building platforms to address these concepts. "These platforms are having to handle privacy, security, tracking and so on; things that we didn't run into before," Green admitted. "We've gone from M2M being about extending ERP systems to machines, to the IoT, in which we are connecting something to the internet." The bottom line, he contended, is that the ways in which we define 'things' has changed. "It's a philosophical change," he believes. "It's exciting, but there is a lot of confusion." Part of that confusion relates to the application of trust, security and authentication. Marini believes that, from an overall system perspective, security and trust are basic requirements. "We need both and both are equally important," she said. Green said trust, security and authentication are culturally dependent. "One person's security is another's infringement and many factors come into play." Marini added: "When you look at industrial or medical systems – even some consumer applications – it's critical to know whether the device is the one to which you should be talking and whether it belongs to the person you think it belongs to. It's an important concept and relies on the development and deployment of robust authentication techniques." Green believes IoT systems will need contextual input. He used his example of the toaster to explain. "Context seems to apply to most things I care about. The fact that I am or am not making toast can be interpreted differently. I may not be making toast because I'm on holiday, but for older toaster users, there may be another reason. "So you have to ask 'what do you mean?'. You have to communicate a context to people accessing information. But then you have to ask whether there are any laws that apply – and European legislation on this aspect is in a tangle." The scale of the problem becomes apparent when the number of connected devices is examined. Depending upon whose research you read, there could be up to 50billion connected devices by the end of this decade. But what are these devices? "That just relates to M2M," Green contended. "In my opinion, the IoT will be at least 100 times that size. We haven't seen anything like its potential scope yet and I anticipate trillions of devices." How does he arrive at that figure? "M2M has been worked out on the basis of 10 connected devices per head. When you think about mobile phone, tablet and so on, it doesn't take you long to get there. But there's also the working environment; the water company has stuff in the street and there's a street lamp outside your house which you can be associated with. "But that's just M2M. With the IoT, it will be things like a lawn mower which communicates with its manufacturer without you knowing that it's doing so. There will be a lot of things talking to each other and how many times will we be involved? If it's working, we won't want to be." Step forward trust, security and authentication procedures. But what are they and where are they coming from? "We're working on this with several partners," Green noted, "but these won't come from one place. If I am going to authenticate a sensor, for example, I will need to know that it comes from a reputable manufacturer and that it has been tested and approved." He says middleware will be the key to making this approach work. "But no approach has, so far, satisfied the end to end problem." At the MCU end, he says companies like Freescale have done well. "But that only goes so far. What is needed is an open standard, but even if that can be developed, it will probably be inappropriate because it will be so vast there won't be any standardisation." Davidson pointed to the human element. "One of the things which people worry about is how security keys are handled. Humans make the process insecure and it's probably something that could be anonymised. But how that works and where you turn on the security is a bigger challenge." What role can ARM play in this? Marini said ARM is looking at the different requirements and how it can connect the different partners in it ecosystem. "It's not about products," she said, "it's about ways of working with partners. This has been a big focus over the last year, looking to create a system that allows every end point to do some encryption, key exchange and device management." The MCUs required in these end devices may well be more powerful than would be expected, but will still need to be cost effective. Marini said common requirements would be low cost end points, some more intelligent devices that can run a full operating system, device management and secure data transport. Freescale, for example, has identified the potential to deploy application processors, rather than MCUs. Geoff Lees, general manager of Freescale's microcontroller business, said: "We are beginning to see more IoT applications suited to low end Cortex-A processors. Application processors are coming to the domestic market – and that's something that was unexpected." Davidson gave more details. "We have been addressing security and trust for some time and have a number of approaches. One is to secure communications using, for example, IPSec or SSL. But we're also looking at things like system level memory management and secure boot. We also have a trust architecture that is an extension of ARM's TrustZone, which protects the CPU and the cache memory subsystem. Alongside software tamper detect, we also offer physical anti tampering mechanisms. "One you've authorised software, you can then allow access to hidden registers and the device can be part of an end to end trusted system. We're talking to a range of customers and technology companies about how they can use these features." Marini added: "End devices are going to need to support things like authentication and over the air updates; things people don't associate with small MCUs." Davidson pointed to the connected car as a potential trouble spot. "One big element of the business case for connected cars is the ability to perform over the air updates. But this will need end nodes with bigger stacks, more memory and higher core performance. That has a cost and the devices draw more power; all challenges which are being worked on." While Green suggests authentication will start at the end device and work its way towards a gateway to the cloud, he pointed out some potential problems. "An MCU needs to come alive, but how does it say 'I'm here'? That's where companies like Freescale come in," he concluded. "But what happens when you build that MCU into something else and that gets built into something else?"