'Twas Brillo but then Android Things, which watched as Google Weaved its Nest Google has launched the developer preview of Android Things, updating and rebranding the Brillo IoT operating system which was unveiled over a year ago. Designed for medium-complexity devices, like home hubs, thermostats, and security cameras, the stripped-down version of Android has had zero success in the market. The new …
Well Google do a pretty good job of supporting Nexus devices directly from Google, so why would this be any different?
I have used the Dev Preview 1 of Android Things, and it's pretty darn good, and a huge step forward for bringing some standardization to IOT. One particular app I already had that was really well suited to IOT compiled easily using Android Studio and a couple of changes to the manifest.
Linux should be kept away from security. Contrary to what most people think, security is most important on end-user and IoT devices. These are uncontrolled devices. On servers, security is not so much of an issue on servers (although maybe more important), since servers are very controlled and behind physically locked doors.
The tradeoff is performance. Security gives a massive performance hit. That is significant on servers trying to serve 1,000s of users at once. But on end-user and IoT devices doing one specific thing, performance is not so important.
Linux trades off security for performance and is thus not so desirable out there in the wilds (at the edge) of the Internet.
Well, 'common-sense' you really don't know what you are talking about, and put up nothing against what I have said.
It is a fact that Linux trades off security for performance.
Security is very important in data centres and servers, but these environments are very well controlled by experts who only install new software and versions after planning and then testing. Thus they are less dependent on the OS being strong.
End users - at the edge of the Internet - don't plan or test before installing new software. That software can come packaged with Malware. In Linux it is easier to get into other processes, since IPC is fast and direct, but not brokered through a Microkernel.
Now, go away and do some thinking before you come and call what I said claptrap.
My first reaction to what you said was also "claptrap", but, on reflection, I think it lacks context. You're saying Linux is insecure and unsuited for the challenges in, "the wilds (at the edge) of the Internet"? So what alternatives have seen massive deployment in that area? Perhaps Windows, iOS, OS X? They certainly don't represent a different level of security, so I'm missing your meaning.
I took a look at your previous posts... are you thinking more in terms of formerly-defined systems, with provable behaviour? In that case I have a few criticisms:
i) Even "simple" IoT devices are too complex for easy formal definition
ii) Your concept of "security" is restricted to technical considerations. Security should encompass confidentiality, integrity and availability, and the trade-off between them is determined by the application.
iii) You say that software can come packaged with malware... but how do you define malware? I tend to use the definition, "software that does bad things", but that requires assessment of intent, which is a human quality not amenable to formal definition. Actually, this is the flip side of my point (ii)...
iv) Costs are being driven down, we don't have very controlled servers in locked data centres (and even when we do, someone's fitted an IoT lock, "for efficiency"), tended by perfect experts.
So, take a look around the real world, it is more complicated and messy than a formally-defined microkernel can cope with.
Allan George Dyer
I thought parts of your answer were good, but you ended with the old 'real world' chestnut. That is exactly what we are talking about.
So, here are some responses.
Yes, iOS and MacOS (formerly OS X) are more secure being based on Mach. See my explanation of how IPC works in these different environments of micro-kernel and non-mk.
CIA certainly exists at higher levels of abstraction and security must be applied at all levels. However, when your lowest level is insecure, all higher levels are affected. Security must be built into systems, not bolted on as an afterthought. However, the discussion was about microkernels, which is technical. So your criticism point ii really amounts to nothing, and really that means not your point about context either.
iii I think we know what malware is - how does giving a 'precise' definition here help? It doesn't. OK, malware is software that has been installed on your machine for nefarious means. As I said, data centres and servers are carefully managed, end user systems not (mostly). Thus the more sensible tradeoff is performance for security, not the Linux tradeoff which is the other way.
iv well most servers (at least ones of any scale) are set up by experts - I never applied the adjective 'perfect'. No, even 'experts' need to review their security constantly.
Of course, there is always more to security, but every part must be right - our discussions was about microkernels.
Sorry about the long delay - I only looked back at this thread today.
Responding to your responses...
iOS and MacOS are based on Mach, but they aren't microkernels: "However, in OS X, Mach is linked with other kernel components into a single kernel address space. This is primarily for performance" https://developer.apple.com/library/content/documentation/Darwin/Conceptual/KernelProgramming/Mach/Mach.html
How this affects their security compared to Linux, I don't know.
Were we restricting the discussion to "how microkernels make IoT secure", or "how to make IoT secure"? My point is that it isn't all about microkernels. Yes, insecurity at a lower level affects all higher levels, but it is also possible to "bolt on" insecurity, To steal an analogy from cryptography, you need strong algorithms and good key lengths, but if your crypto is like a 1 mile stake in the ground - infeasible to get through or over, increasing the key length to make it a 10 mile stake doesn't improve your security - the attackers still go around the stake instead. Securing your IPC messages against an obscure attack doesn't fix a hole elsewhere. As an example of "bolt on" insecurity, SMTP email was a mostly harmless protocol until someone (Microsoft) decided to add a programming language (VBS) to their email client, and have it autorun scripts in messages on preview... then we got Loveletter. All the inherent insecurities of the underlying protocol made no difference to how the catastrophe spread, it was the add-on scripting that made the difference.
How does a precise definition of malware help? I'm saying we can't really have one, but the idea of securing everything from the microkernel depends on it. We can't really have one because the classification depends on intent: format is a useful program, but a trojan with the same function is highly destructive. So, we'll ask the microkernel, "which of these two programs that overwrite the disc should be allowed to run?", perhaps it will refer to a signature: who wrote the program? Then, who do we trust? Has the key been compromised? We're back to dancing on quicksand.
"well most servers (at least ones of any scale) are set up by experts" - Then why do we see news stories like https://www.theregister.co.uk/2016/09/22/yahoo_500m_email_accounts_hacked/ ? I would say that most servers are secured to a cost, and an expert will make a rational choice between meeting an unreasonable constraint and loosing their job.
Spend too much on a secure microkernel, and there's not enough left to identify or fix gaping holes elsewhere.
Very simplistic opinion from somone that hasn't saved money from nest. Mine is about the only investment that has paid for itself in 2 years.
I save about a tenner a month on my heating compared to my previous dumb thermostat that worked as you describe. My house also feels warmer too and it's warm at the right times without needing to tinker with settings.
It's way smarter that you imagine, combining data from historic behaviour patterns, current outside temperature, forcasts of weather, time to heat, time to cool and more.
A very smart device that saves real money
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
