Apple's Strict Bluetooth LE Security Requirements Slowing Rollout of HomeKit Accessories

[prasand]

Security first absolutely, but did you read the insane requirements?

Elliptic Curve and 3072 bit keys... What are we guarding against, quantum computers? What's wrong with good old AES 256?

AES 256 is vulnerable. Granted, proper implementations have proven to be uncrackable without significant cryptanalysis ... but many do not use adequate implementations / algorithms and it's relatively simple to attack / break that weak link without being concerned with the size of 256-bit key lengths.

3072-bits might be a 'bit' excessive, but I don't think things like ECDSA are (Elliptic Curve Digital Signature Algorithm). Considering, the successful attack on AES 256 is more than a few years old and is actively being worked on, a practical break could occur any day (it's highly possible that one already privately exists). But 'when' that indeed does happen, a simple firmware update would not address the problem when already installed hardware cannot handle the switch.

Personally, I would've said 1024-bits minimum but the hardware must be able to handle 2048-bits.

-- Edit --

https://en.wikipedia.org/wiki/Biclique_attack

 

[57004]

Does this Encryption requirement apply to ALL Bluetooth LE connections or just when using HomeKit?

I'm considering developing an app (and some open source hardware) which will not have any need for privacy and will not be using HomeKit. Basically it will interface to a Ham radio (which is public anyway).

I'm just wondering if I would need to do Encryption to get this to work - I hope not as it would make the project unnecessarily difficult. I've developed Apps before but never done anything with Bluetooth LE.

 

[prasand]

Does this Encryption requirement apply to ALL Bluetooth LE connections or just when using HomeKit?

It only pertains to HomeKit. You're safe / okay.

 

[charlituna]

Security first... efficiency later.

I'm okay with that.....

presumably the strict security is to make some some yahoo can't bust my smart lock and get into my place. so yeah I'm good with that

 

[Benjamin Frost]

Dear oh dear.

40 seconds. Pfft.

HomeKit is never going to take off. It's dead in the water.

 

[unplugme71]

Translation of the article: It will be 2 to 3 years minimum before HomeKit becomes a viable technology for the vast majority of Apple's customer base. Forty seconds to unlock a door? The average person is going to say "screw that" and just continue to use a key. If my brand new Mac took 40 seconds to boot, that would be considered a problem.

Truth be told, Apple should've been talking to chip makers and app developers behind the scenes for at least another year or two before they announced HomeKit to the public. This "just get something shipped and we'll fix it later but we're going to present this new thing at the keynote as if it's ready for prime time and it's groundbreaking" approach that Apple seems to be taking more and more these days is getting old.

Horrible idea. Better to work slow like Apple TV and release things but by bit. Yes some might get screwed over when HK 2.0 arrives and some devices won't work but it'll help shape HK

 

[unplugme71]

All that does is annoy neighbours and anyone walking past, neither of which comes running to help or see what the commotion is.

No one is dumb enough to potentially be killed. I would look out the window, grab my red dot laser, put it right in front of him and say wait for the police you ass hat.

 

[Solomani]

This is one of those forward thinking things that will inconvenience us now but be worthwhile later.

No. This will not be tolerated when the Zombie Apocalypse comes. Imagine that a zombie horde is in your front yard, and you sprint to the front door to electronically lock it (from inside) via HomeKit device.

"Countdown to Door Lock…. 40 seconds…. 39 seconds…. 38 seconds….."

The zombies would have gotten inside your house before the electronic door would have finished its lockdown sequence. The feasting on brains is guaranteed.

 

[dwaltwhit]

Maybe I'm missing something but why is there not a collection of Apple components for this? Wh is apple only relying on third parties? Isnt the allure of apple the fact that the hardware is made for the software and vice versa? Why don't we have apple locks? Apple thermostats? Switches?

 

[Zxxv]

No one is dumb enough to potentially be killed. I would look out the window, grab my red dot laser, put it right in front of him and say wait for the police you ass hat.

In the UK people don't even do that. Like I said it's just an annoying noise. No ones coming to see except the private company you paid and they will take awhile.

 

[Cuno]

Security first absolutely, but did you read the insane requirements?

Elliptic Curve and 3072 bit keys... What are we guarding against, quantum computers?

Given the long-term investments for home automation equipment, and the arms race for better crypto algorithms and better ways of breaking them, these requirements are not at all insane. They are simply state of the art.

Moreover, 40 seconds for opening a connection and reading the door lock state indicates a non-optimized, first generation implementation of the HomeKit Accessory Protocol. Our highly optimized third-generation implementation (http://oberonhap.com) takes 1.1 seconds for the crypto processing for opening a connection on a relatively slow 16 MHz Nordic nRF51, and negligible time on the newer nRF52. Also, we brought down RAM consumption to less than 12 KB. This is not insane even for low cost, mass market devices.

There's not even a need to throw more hardware after this software problem. Solid software engineering is perfectly sufficient.

 

[TheRealTVGuy]

I'm just patiently waiting for the day when I can lock my front door with my AppleWatch, go for a run with just the watch and some bluetooth headphones, then once I arrive home, unlock the door with the watch.

 

[longofest]

But this gear isn't like buying an iPhone that I'm going to toss in two years. If I install locking and lighting and heating systems deeply embedded into my house, I'm going to want at least some of that stuff to last thirty years. Folks are installing solar panels, LED lighting, boilers etc. which don't pay for themselves for at least 10, 15, 20 years. Whatever protocols they design now, really have to last for a very very long time. Home automation isn't going to work if Apple pull their usual stunt of forcing us to upgrade all our hardware on their schedule. Futureproofing their security is a promising suggestion that they won't.

No encryption or key length is guaranteed to last 30 years. When I installed my wireless lock in my door, I absolutely don't have the expectation that it will last that long. 3DES was gold standard when it was published in 1998, I'd say by 2008 it was already being supplanted by AES as the chosen cipher by most folks.

Honestly, I take a bit of issue with the idea of "futureproofing security". There really isn't any such thing. Pick a modern and well-accepted form of security, and then have a method to apply patches.

 

[Cuno]

No encryption or key length is guaranteed to last 30 years. When
Honestly, I take a bit of issue with the idea of "futureproofing security". There really isn't any such thing.

That's true indeed. However, about 10 years ago, the elliptic curve algorithms of Bernstein et.al. have begun to supplant older algorithms, with Apple, Google and others using them in various places.
The choice of SRP is a bit conservative though, here a more efficient and similarly secure choice might have been possible, but at the cost of inventing something new. Which, in cryptography, most people and companies should never attempt. Fortunately, with regard to speed, SRP is only used once, during the initial setup of a device.

 

[longofest]

Given the long-term investments for home automation equipment, and the arms race for better crypto algorithms and better ways of breaking them, these requirements are not at all insane. They are simply state of the art.

Moreover, 40 seconds for opening a connection and reading the door lock state indicates a non-optimized, first generation implementation of the HomeKit Accessory Protocol. Our highly optimized third-generation implementation (http://oberonhap.com) takes 1.1 seconds for the crypto processing for opening a connection on a relatively slow 16 MHz Nordic nRF51, and negligible time on the newer nRF52. Also, we brought down RAM consumption to less than 12 KB. This is not insane even for low cost, mass market devices.

There's not even a need to throw more hardware after this software problem. Solid software engineering is perfectly sufficient.

See my latest post. I'll give kudos to you for what you did with your implementation (I looked through your website a bit), but at the same time, you are still at least dealing with an ARM based platform. I know guys who are trying to do this stuff using Atmel AVR based stuff.

 

[Cuno]

I know guys who are trying to do this stuff using Atmel AVR based stuff.

Yes, that's a tough challenge ;-) If feasible, the results may not be pretty.

 

[Cuno]

I know guys who are trying to do this stuff using Atmel AVR based stuff.

We haven't studied the AVR instruction set with regard to HomeKit yet, but maybe it would be worth a try. We love challenges...

 

[MarkCollette]

Security first absolutely, but did you read the insane requirements?

Elliptic Curve and 3072 bit keys... What are we guarding against, quantum computers? What's wrong with good old AES 256?

Appliances are likely to be used for decades, longer time-frames than crypto algorithms remain secure, hence using something well forward oriented.

 

[iConnected]

Dear oh dear.

40 seconds. Pfft.

HomeKit is never going to take off. It's dead in the water.

Apple according to Benjamin Frost:

HomeKit: "never going to take off", "dead in the water".
Apple Watch: "flawed", "beta product", "useless", "pointless", "flop", "not a category worth pursuing".
Apple Music: "flop", "Crud 1 should be the name of their junk radio station".
Beats Solo2 headphones: "flopping".
Tim Cook: "failed", "Someone please sack the man".
Apple car: jokes about a "Titanic project".
Apple: "unfocused", "beta software for everyone", "corruption ... goes unrecognised for a while ... So it is with Apple.", "idiots", "monstrous evil dictatorship".
iPhone: "overpriced".
iPad mini 3: "rip-off".
Apple Pay: "not sure that it's worthwhile for Apple".
iPad: "abandoned".
iPad Pro: "Great for sheet music".
iPhone 6 Plus / phablets: "stupid".
Sir Jonathan Ive: "gone off the boil in recent years".
MacBook: "ridiculously compromised".
Apple TV: "neglected hobby".

There seems to be a common theme.

 

[prasand]

No encryption or key length is guaranteed to last 30 years.

I know guys who are trying to do this stuff using Atmel AVR based stuff.

Indeed. I think Rijndael / SHA-2 is relatively on it's way out the door. Not just from a private vulnerability standpoint but the fact that Keccak (SHA-3) is actually faster and more efficient. I believe the former has out-lived its purpose, as it was not intended for this mobile era nor suitable for the next (wearables or implants). But until SHA-2 is actually supplanted, your latter reply is exactly why Apple needs to force manufacturers to up their game.

Though personally, Blake2 has my heart. Color me "knife to eye" rebellious.

Maybe I'm missing something but why is there not a collection of Apple components for this? Wh is apple only relying on third parties? Isnt the allure of apple the fact that the hardware is made for the software and vice versa? Why don't we have apple locks? Apple thermostats? Switches?

I doubt that Apple is relying on third parties. Some items might be too banal for Apple to pursue, since there's only but so much one can do with a light switch, lock, or a garage door opener. But for center-piece devices such as a thermostat, the third party solutions aren't compelling in an Apple aesthetic nor logic. So either Apple has already begun making moves to create its own solution and are using the 3rd party to fill the time-gap, or they are in a "wait and see" stance and will create their own solution if the 3rd parties do not come up with something compelling in the short-term. Apple TVs aren't meant to be on coffee tables, so they are not a conversation piece. Apple "not" doing a conversation piece for HomeKit, I don't really see. Something like a thermostat is usually in the center of the home (psychological center / living room), in or near the location most entertain guests, or is one of the first things people see upon entering their home. I think it's only a matter of time before Apple ceases to give up that prime brand identity space to third parties.

At least one thing that would cause most to associate a location as an "Apple Home", I imagine Apple will pursue. Be that a thermostat or something(s) else.

 

[Zxxv]

There's not even a need to throw more hardware after this software problem. Solid software engineering is perfectly sufficient.

seeing how crap the philips hue app is I'm sure it will be a few years before the philips team realise they need to hire someone else

 

[Zxxv]

Maybe I'm missing something but why is there not a collection of Apple components for this? Wh is apple only relying on third parties? Isnt the allure of apple the fact that the hardware is made for the software and vice versa? Why don't we have apple locks? Apple thermostats? Switches?

apple makes 30% off app purchases

 

[Cuno]

Though personally, Blake2 has my heart.

+1

 

[rhett7660]

Having automation in the home which could open it up to those who are going to try and use it to get information or get inside of your house is huge. I would rather, wait and have a product that is delivered that is pretty darn secure.