Understanding MongoDB's Object ID

Bharat Kalluri / 2020-10-01

MongoDB ID's are not entirely random. The actual spec is in the mongoDB documentation here.

The gist of it is that, it is a 12 byte value in which

  • The first 4 bytes represent the ObjectID creation epoch timestamp
  • 5 bytes random value
  • 3 byte incrementing counter, initialized to a random value

For example, 5f6f613ad7bf070237959759 is a valid mongoDB ID.

mongo_id = "5f6f613ad7bf070237959759"
ts_hex = mongo_id[:8]
ts = int(ts_hex, 16)
# 1601134906 => Saturday, September 26, 2020 3:41:46 PM GMT

The next 5 bytes (mongo_id[8:18]) are random, so nothing to dig in there. The next 3 bytes are an incrementing counter, but initialized at a random value. Consider two ID's 5f70d73ae16328c827714546 and 5f70d741e16328c827714547, the last 3 bytes are incremented (714546 and 714547). This is one of the reason mongoID's are sortable, since they have the timestamp and an auto incrementing key in them already.

But objectID's created at the same time do not ensure ordering. And one more piece to note is that the developer can also generate object ID's. And if the program runs in two different computers, they can have their system clocks at different positions.

The interesting history of MongoDB ID formats

In version 3.2 (and before afaik), ObjectID's had a different format. They did not have the 5 byte random value., This made it possibly vunerable to IDOR (Insecure Direct Object Reference) attacks if the developer did not put enough thought into the system. Post 3.2 the entropy was introduced to make sure people cannot guess the next identifier.

Hand crafted by Bharat Kalluri