Skip to content

Hybrid cryptography explained, classical and post-quantum side by side

A hybrid construction combines a classical algorithm with a post-quantum one so that the result is secure if either holds. X25519MLKEM768 is the common form in TLS today. It exists because post-quantum algorithms are new, and it costs bandwidth rather than confidence.

Why hybrid at all

Lattice cryptography has been studied for decades but the standardised parameter sets are recent, and the implementations more recent still. Classical elliptic curve has decades of deployment and cryptanalysis behind it — and a known quantum weakness.

Combining them means an attacker must break both. A cryptanalytic advance against lattices leaves X25519 protecting the session; a quantum computer leaves ML-KEM protecting it.

How it works in TLS

Both key exchanges are performed and their outputs combined into a single shared secret through a key derivation function. The construction is arranged so the result is at least as strong as the stronger component.

In practice this appears as a named group — X25519MLKEM768 — which browsers and servers negotiate like any other. Deployment has been rapid, and for most operators it has already happened without a project.

What it costs

Handshake size, mainly. Carrying both a classical and a post-quantum key exchange makes the ClientHello substantially larger, which can push it across packet boundaries and has caused real interoperability problems with middleboxes that made assumptions about handshake size.

Computation is a minor concern. The bandwidth and the middlebox behaviour are the practical issues, and both are worth testing rather than assuming.

How long it persists

Longer than the enthusiasm for removing it. Hybrid modes stay until confidence in the post-quantum component is high enough that the classical half is redundant, and that is a judgement different authorities will make at different times.

Planning for hybrid to be a decade-long state rather than a transitional one is the safer assumption. Which is another instance of the same lesson: whatever you deploy has to be able to change again — agility rather than a single migration.

The entropy consequence

Hybrid means performing two key exchanges, each consuming randomness. Entropy demand per handshake goes up rather than down during the transition, which is worth including in the sizing calculation.