What is a signed NDEF record, and what does it prove?
That the content carries a signature made with a particular key, and that the content has not changed since. What it does not establish is that this tag is the original, because a signature travels with the content and copies perfectly onto a blank tag alongside everything else.
What it is
A record carried alongside the others in the message, holding a signature computed over them and a way of identifying the key that produced it.
It is content, in exactly the sense that everything else on the tag is content. It occupies space in the message and is read like any other record.
The record type, the algorithms permitted and the certificate arrangements are defined by the Signature Record Type Definition, an NFC Forum technical specification. It deliberately does not define a public key infrastructure, which is the gap this page describes.
Which platforms and applications verify it, rather than merely carrying it, is the question that decides whether it does anything in practice.
What it proves
That whoever prepared this content held the key. If the key is genuinely private to a publisher, then content bearing their signature came from them.
And that the content has not been edited since. A record altered after signing no longer matches, so the signature fails.
Those two properties are real and are worth something. The trouble begins when people extend them to a third property that does not follow.
What it does not prove
That this tag is the tag the publisher wrote. The signature is inside the content, the content copies completely, and the copy carries a signature that verifies perfectly.
This is the point people find counterintuitive and it is the whole of the matter: a signature authenticates a message, not the object carrying it.
It also proves nothing about the item the tag is stuck to, and nothing about whether the tag has been moved from one item to another.
For any of those, the proof has to come from the chip rather than from the content, which is a different mechanism with a different cost.
Where it is worth using
Where the concern is altered content rather than duplicated tags: a tag on a public wall that somebody might rewrite to point elsewhere.
Where a reader on the other side actually checks it, which usually means a purpose built application rather than the ordinary system behaviour.
It costs memory, which on a small chip is a real constraint, and it costs the operational work of holding a key safely for years.
Where the concern is counterfeiting, it is the wrong tool and buying it will produce a scheme that fails exactly when it is tested.
Frequently asked questions
Does a signature stop a tag being copied?
No. The signature is content, and content copies completely. The copy carries a signature that verifies.
Does it stop content being altered?
Yes, in the sense that altered content no longer matches the signature. That is what it is for.
Will a phone check it automatically?
Generally not. Verification usually needs an application built to do it, which is what decides whether it helps.
What should I use for counterfeiting instead?
A chip that proves it holds a key. The proof has to come from the chip rather than from the content.