Understanding Device IDs¶
Every device is identified by a device ID. The device ID is used for address resolution, authentication and authorization. The term “device ID” could interchangeably have been “key ID” since the device ID is a direct property of the public key in use.
To understand device IDs we need to look at the underlying mechanisms. At first startup, Syncthing will create a public/private keypair.
Currently this is a 384 bit ECDSA key (3072 bit RSA prior to v0.12.5,
which is what is used as an example in this article). The keys are saved in
the form of the private key (
key.pem) and a self signed certificate
cert.pem). The self signing part doesn’t actually add any security or
functionality as far as Syncthing is concerned but it enables the use of the
keys in a standard TLS exchange.
The typical certificate will look something like this, inspected with
Version: 3 (0x2)
Serial Number: 0 (0x0)
Signature Algorithm: sha1WithRSAEncryption
Not Before: Mar 30 21:10:52 2014 GMT
Not After : Dec 31 23:59:59 2049 GMT
Subject Public Key Info:
Public Key Algorithm: rsaEncryption
RSA Public Key: (3072 bit)
Modulus (3072 bit):
Exponent: 65537 (0x10001)
X509v3 Key Usage: critical
Digital Signature, Key Encipherment
X509v3 Extended Key Usage:
TLS Web Server Authentication, TLS Web Client Authentication
X509v3 Basic Constraints: critical
Signature Algorithm: sha1WithRSAEncryption
We can see here that the certificate is little more than a container for the public key; the serial number is zero and the Issuer and Subject are both “syncthing” where a qualified name might otherwise be expected.
An advanced user could replace the
cert.pem files with a
keypair generated directly by the
openssl utility or other mechanism.
To form a device ID the SHA-256 hash of the certificate data in DER form is
calculated. This means the hash covers all information under the
Certificate: section above.
The hashing results in a 256 bit hash which we encode using base32. Base32 encodes five bits per character so we need 256 / 5 = 51.2 characters to encode the device ID. This becomes 52 characters in practice, but 52 characters of base32 would decode to 260 bits which is not a whole number of bytes. The base32 encoding adds padding to 280 bits (the next multiple of both 5 and 8 bits) so the resulting ID looks something like:
The padding (
====) is stripped away, the device ID split into four
groups, and check
are added for each group. For presentation purposes the device ID is
grouped with dashes, resulting in the final value:
Now we know what device IDs are, here’s how they are used in Syncthing. When you add a device ID to the configuration, Syncthing will attempt to connect to that device. The first thing we need to do is figure out the IP and port to connect to. There are three possibilities here:
The IP and port can be set statically in the configuration. The IP can equally well be a host name, so if you have a static IP or a dynamic DNS setup this might be a good option.
Using local discovery, if enabled. Every Syncthing instance on a LAN periodically broadcasts information about itself (device ID, address, port number). If we’ve seen one of these broadcasts for a given device ID that’s where we try to connect.
Using global discovery, if enabled. Every Syncthing instance announces itself to the global discovery service (device ID and external port number - the internal address is not announced to the global server). If we don’t have a static address and haven’t seen any local announcements the global discovery server will be queried for an address.
Once we have an address and port a TCP connection is established and a TLS handshake performed. As part of the handshake both devices present their certificates. Once the handshake has completed and the peer certificate is known, the following steps are performed:
Calculate the remote device ID by processing the received certificate as above.
Weed out a few possible misconfigurations - i.e. if the device ID is that of the local device or of a device we already have an active connection to. Drop the connection in these cases.
Verify the remote device ID against the configuration. If it is not a device ID we are expecting to talk to, drop the connection.
Verify the certificate
CommonNameagainst the configuration. By default, we expect it to be
syncthing, but when using custom certificates this can be changed.
If everything checks out so far, accept the connection.
An Aside About Collisions¶
The SHA-256 hash is cryptographically collision resistant. This means that there is no way that we know of to create two different messages with the same hash.
You can argue that of course there are collisions - there’s an infinite amount of inputs and a finite amount of outputs - so by definition there are infinitely many messages that result in the same hash.
I’m going to quote stack overflow here:
The usual answer goes thus: what is the probability that a rogue asteroid crashes on Earth within the next second, obliterating civilization-as-we- know-it, and killing off a few billion people ? It can be argued that any unlucky event with a probability lower than that is not actually very important.
If we have a “perfect” hash function with output size n, and we have p messages to hash (individual message length is not important), then probability of collision is about p2/2n+1 (this is an approximation which is valid for “small” p, i.e. substantially smaller than 2n/2). For instance, with SHA-256 (n=256) and one billion messages (p=10^9) then the probability is about 4.3*10^-60.
A mass-murderer space rock happens about once every 30 million years on average. This leads to a probability of such an event occurring in the next second to about 10^-15. That’s 45 orders of magnitude more probable than the SHA-256 collision. Briefly stated, if you find SHA-256 collisions scary then your priorities are wrong.
It’s also worth noting that the property of SHA-256 that we are using is not simply collision resistance but resistance to a preimage attack, i.e. even if you can find two messages that result in a hash collision that doesn’t help you attack Syncthing (or TLS in general). You need to create a message that hashes to exactly the hash that my certificate already has or you won’t get in.
Note also that it’s not good enough to find a random blob of bits that happen to have the same hash as my certificate. You need to create a valid DER-encoded, signed certificate that has the same hash as mine. The difficulty of this is staggeringly far beyond the already staggering difficulty of finding a SHA-256 collision.
Problems and Vulnerabilities¶
As far as I know, these are the issues or potential issues with the above mechanism.
Currently, the local discovery mechanism isn’t protected by crypto. This means that any device can in theory announce itself for any device ID and potentially receive connections for that device from the local network.
Long Device IDs are Painful¶
It’s a mouthful to read over the phone, annoying to type into an SMS or even into a computer. And it needs to be done twice, once for each side.
This isn’t a vulnerability as such, but a user experience problem. There are various possible solutions:
Use shorter device IDs with verification based on the full ID (“You entered MFZWI3; I found and connected to a device with the ID MFZWI3-DBONSG-YYLTMR-WGC43E-NRQXGZ-DMMFZW-I3DBON-SGYYLT-MRWA, please confirm that this is correct”).
Use shorter device IDs with an out of band authentication, a la Bluetooth pairing. You enter a one time PIN into Syncthing and give that PIN plus a short device ID to another user. On initial connect, both sides verify that the other knows the correct PIN before accepting the connection.