th0mas
commented
4 years ago From the (small) amount of research I've done the general consensus is that Bluetooth 4 / Bluetooth low energy is "secure". However i'm not sure how well Bluetooth stands up to replay and man-in-the-middle attacks.
I think by treating Bluetooth as an insecure protocol we gain several benefits:
- Protects against future attacks on Bluetooth.
- Protects against our own mistakes configuring and authenticating - Bluetooth is a complex protocol!
- Allows for use of other protocols in future devices such as Zigbee.
- Allows users to use multiple devices if necessary, such as a phone and smartwatch.
I believe we could easily implement a challenge response protocol over Bluetooth that would be plenty secure enough for use on a door lock.
Sharing private keys in a initial setup phase could easily be done over WiFi or, if needed, the wider internet over TLS.
Challenge-Response
- App asks the door to unlock over BLE
- The door responds with a randomly generated number
- The App signs this number with a shared/private key
- The door then verifies the signature before unlocking
This may be vulnerable to Relay attacks, but this could be pretty easily mitigated by adding a timestamp to the signed response number, making it impossible to use stored responses at a later date.
There is an industry standard protocol based on this, Kerberos, but it looks a bit heavy handed for our usage.
Other solution
This is arguably less secure, but is simpler (less to go wrong) and plenty "good enough" - especially if we presume Bluetooth has good security on top of this.
- App signs the current time and date and sends an unlock command over BLE.
- Door checks signature and makes sure time is within bounds before unlocking.
This however, won't work if the door does not have a reliable way to keep time.
The running theme of the research on "Smart Door Locks" I've done so far is the biggest vulnerability is the locks themselves to "low tech" attacks. Some of the most cryptographically secure locks on the market are vulnerable to simple attacks with screwdrivers.
Remote entry is a hard problem, Tesla, VW, Citroen, Ford and Peugeot have struggled to use an implement secure systems.
Its worth remembering that most physical locks are also vulnerable to picking and other physical attacks, so your not gaining or loosing any "real world" security by making a well-implemented smart lock.
Some interesting reading on remote access systems, some aren't Bluetooth based but still give a good overview of why remote locks are so hard to do properly.
https://morphuslabs.com/hacking-the-nok%C4%93-padlock-adfe7b1b5617 - Hacking a BLE padlock - used a hardcoded key however so this shouldn't apply to a challenge-response solution.
https://www.cs.bham.ac.uk/~oswalddf/publications/2016_usenix_lock_it_and_still_lose_it.pdf - A research paper on various car keys.
https://eprint.iacr.org/2010/332.pdf - Another research paper targeting car keys.
nelsonic
We are planning to use Bluetooth Low Energy (BTLE) to authenticate people via their Phone (or other BTLE capable device) in order to enter the house
@home.We know this is how the keyless entry works on Tesla Model 3: https://www.tesla.com/support/car-safety-security-features
We need to investigate if the Bluetooth Identifier for the device can bee spoofed or if there is any way to crack the data transmission. We are going to be entrusting a lot of value into this system so we need to know that the foundations we are building on are rock solid.
So far this is what I've read:
Is Bluetooth 4.0 traffic encrypted by default/design? https://security.stackexchange.com/questions/100554/is-bluetooth-4-0-traffic-encrypted-by-default-design
Is it possible to spoof a paired Bluetooth device? https://security.stackexchange.com/questions/139854/is-it-possible-to-spoof-a-paired-bluetooth-device
Todo