ML-DSA vs SLH-DSA: Signature Size and Verify Cost in Post-Quantum Deployments
A machine that runs Shor's algorithm would break the signatures holding up most of the internet, and two families are competing to replace them. Module-lattice ML-DSA signatures rest on structured lattice problems and ke
A machine that runs Shor's algorithm would break the signatures holding up most of the internet, and two families are competing to replace them. Module-lattice ML-DSA signatures rest on structured lattice problems and keep keys and signatures compact. Stateless hash-based SLH-DSA signatures rest only on hash functions, the most conservative assumption available, and pay for that caution in bytes and signing time. The choice looks like a security argument; it is really a bandwidth and latency argument.
ML-DSA-65 produces a 3.3-kilobyte signature and a 1.9-kilobyte key, while SLH-DSA at the 128-bit small set needs 7.9 kilobytes per signature but only 32 bytes for the key.
Part One: What is being compared
The headline number is the wrong unit: a budget records cost before the first result, per operation, and on bad days.
Both replace elliptic-curve signatures with something a quantum computer cannot invert, and both are standardised and shipping in libraries today.
Part Two: The arithmetic
The arithmetic is dominated by size. ML-DSA-65 signs to roughly 3.3 kilobytes with a 1.9-kilobyte public key, and verification costs tens of microseconds on a server core. SLH-DSA at the 128-bit small parameter set produces signatures near 7.9 kilobytes but a public key of only 32 bytes, and verification runs into hundreds of microseconds because the verifier recomputes a hash chain. Signing is the expensive side for the hash-based family, often milliseconds and thousands of hash calls.
The mechanism is about information: one design makes part of the problem observable, and someone must still calibrate the instrument.
Part Three: Where the model breaks
The break-even depends on what you store and how often you verify. A certificate chain carrying three signatures grows by roughly 10 kilobytes with the hash-based option, invisible in a browser handshake and painful inside a block or a low-power radio packet. Lattice schemes keep the wire format closer to today's curves, but their security rests on structured problems that have had less adversarial study than hash functions.
Hash-based signing costs thousands of hash calls and milliseconds, so the cheapest verifier in the family is also the most expensive issuer.
Part Four: The later cost
There is an operational bill on both sides. Lattice code must be constant-time and must never reuse the randomness that masks the secret; a single repeated nonce has broken real deployments. Hash-based code has a tiny implementation surface and no such trap, but slow signing pushes work onto constrained devices and multiplies the cost of a certificate authority issuing millions of certificates.
Expect a split by scale: large operators amortise the fixed side, smaller ones stay because their load is spiky.
Part Five: Outside the team
For the person holding a phone, the difference shows up as firmware size, battery and whether a secure element can finish a signature before the screen times out. A wallet that adds eight kilobytes to every transaction, or a sensor that spends a second signing one reading, changes what the product can do.
Outside engineering the consequence is access: a system that needs less of the scarce resource can be replicated more widely.
A fair comparison prices the signature at the volume and packet size the system actually uses, including transmission of the public key, and treats the security assumption as a separate line rather than a marketing claim.
Market sizing check
If post-quantum migration reaches most public key infrastructure by 2032, the tooling layer of libraries, hardware modules, certificate authorities and audit services is plausibly $2B to $7B a year. The swing factor is how long signatures must stay valid after issuance.
All figures are a synthesis of published ranges and a stated model, not a measurement of a shipped system.
The comparison that matters
The comparison that matters is between compact signatures on newer mathematics and bulky signatures on the oldest assumption we have. Lattice ML-DSA wins wherever bytes and latency are billed; hash-based SLH-DSA wins where a small trusted implementation matters more than size. Most systems should ship both and choose per channel.
The ranking assumes the workload keeps its shape; shift the mix and the result changes with no technology changing.
Choose per channel: lattice signatures where bandwidth is metered, hash-based signatures where implementation surface and assumption risk dominate.
SOL: HCqPmPJM3NVtS6YrmVaTZYnuP8XZ2PFgQZKT94VK1LkS
If this saved you time, a tip is appreciated.
Originally published by Dev.to Security. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.
