Release 13 and 14: the groundwork the 5G security set was built on

The defining additions in Release 13 were about devices that are not phones. Narrowband and low-complexity machine-type categories were specified to make it practical to attach very large numbers of cheap, low-power devices to a mobile network: meters, sensors, trackers, things expected to run for years on a battery and to send small amounts of data infrequently. The release also added means of using unlicensed spectrum alongside licensed, and further antenna work on the radio side.

None of that is a security feature, and the security specifications of the release are largely continuous with what preceded them. The significance is in what it committed the network to serving. A handset has a power budget that can absorb cryptographic work, an owner who notices when it misbehaves, and a replacement cycle of a few years. The new categories have none of the three. A device that must last a decade on one battery cannot afford expensive protocol exchanges, has nobody watching it, and will still be in service long after the assumptions behind its credentials have expired.

That is the constraint the archive’s own work on group authentication addresses: if authenticating each device individually is too expensive for a population of this size and this power budget, the exchange has to be restructured.

Release 14 added direct vehicular communication, allowing vehicles to exchange messages with each other and with roadside equipment, and extended the broadcast and multicast capabilities. It also carried a substantial body of study work whose output landed in the release that followed, which is the ordinary rhythm of the process: the investigation happens in one release and the specification in the next.

The vehicular work introduced a security problem that does not arise for handsets. A vehicle broadcasting its position and speed to whatever is nearby has to be believable without being identifiable, because a receiver needs to know the message is genuine and a bystander must not be able to follow the vehicle by listening. Those two requirements pull against each other, and the resolution involves rotating credentials instead of a single durable identity. It is the same tension that the subscriber identity presents, in a setting where the answer had to be found immediately.

What neither release did was revisit the disclosure of the permanent subscriber identity. The problem was well documented by this point, the equipment to exploit it was commonplace, and the fourth generation ended with the fallback intact. A fix would have meant changing the exchange that every existing device already implemented, and that is the kind of change a new generation absorbs and a maintenance release does not.

Read forwards from here, the shape of Release 15’s security work is easier to follow. It arrived with a network expected to serve handsets, constrained long-lived devices and vehicles at once, in a core that had been redesigned as a set of services. The identity problem was the outstanding item on a list that the previous two releases had lengthened.

Release 15 is where the identity disclosure, the key hierarchy and the protection of the new interfaces are all addressed together, and where the concealment described on the concealment page is introduced.