Security & Privacy

Signing a PDF: what a signature image proves, and what it does not

Most documents people call "signed" contain a JPEG of a name. That is often fine, and it is worth knowing exactly what it is, because it proves almost nothing on its own.

5 minute read

You are sent a contract, you draw your name with a trackpad, you place it above the line, you send it back. The document is now signed in the sense everyone means in day-to-day work. It is also, technically, a PDF with a small picture of some handwriting in it, and nothing about the file itself resists anyone moving that picture to a different page of a different document.

Whether that matters depends on what you are signing. Knowing the difference is what lets you decide.

Three different things, all called signing

A signature image. A drawn, typed or photographed likeness of your name, placed on the page as an image. It changes how the document looks. It carries no information about who put it there or whether the text above it changed afterwards.

An electronic signature with an audit trail. What DocuSign, Adobe Sign and similar services provide. The visible mark is still a picture, but the service records who accessed the document, from which IP address, at what time, with what email verification, and keeps that record. The evidence lives in the service's logs, not in the file.

A digital signature. A cryptographic operation. The signing software hashes the exact byte range of the document, encrypts that hash with a private key held by the signer, and embeds the result along with a certificate identifying the key holder. A viewer can recompute the hash and check it against the signature.

Only the third one is verifiable from the file alone, with no third party to ask.

What a digital signature actually proves

Two things, and it is worth being precise about both.

Integrity. The bytes covered by the signature have not changed since it was applied. Change one character and the recomputed hash no longer matches, so the viewer reports the signature as invalid. This is the same mechanism behind checksums, applied to a byte range inside a document rather than a whole file.

Authenticity, conditionally. The signature was produced by whoever holds the private key matching the certificate. Whether that person is who the certificate claims depends entirely on the certificate authority that issued it and on the key not having leaked. A self-signed certificate proves the same key signed twice; it proves nothing about identity.

Note what is missing from both. Neither proves the signer read the document, agreed with it, or had authority to commit anyone to it. Those are legal questions and cryptography does not touch them.

Why editing invalidates a signature

People report this as a bug. It is the feature working.

The signature covers a byte range. Rotating a page, adding a page number, merging the file with another, or running it through a compressor all rewrite the file's objects and cross-reference table. The bytes change, the hash changes, the check fails. A signature that survived editing would tell you nothing, because anyone could edit the document and keep the signature attached.

PDF does allow legitimate additions through incremental updates, where new content is appended and the original byte range is left untouched. That is how a second person signs a document the first person already signed, and how form fields get filled after signing if the signer permitted it. Whether such changes are allowed is recorded at signing time. This is also why signing should be the last operation: compress, merge, add page numbers and redact first, then sign the finished article.

The redaction trap

This one causes real harm, so it is worth stating plainly.

Drawing a black rectangle over text in a PDF is the same category of action as pasting a signature image: you have added a graphic on top. The text is still in the content stream underneath. Select it, copy it, paste it into a text editor, and there it is. The same applies to hidden layers and to cropping, which only changes the visible box, not the content outside it.

Proper redaction removes the underlying characters, not just the view of them, and the full explanation is worth reading if you do this often. If you are signing a document you have also redacted, get the redaction right first, because signing a file whose secrets are still in it simply certifies that the secrets are unchanged.

Flatten before you send

If you filled in form fields, the values live in interactive field objects rather than in the page content. They can be edited by the recipient, they can be cleared by a viewer that mishandles forms, and they can collide with field names if the document is merged into another.

Flattening converts those values into ordinary page content. After flattening, what the recipient sees is what the file contains, with no separate layer that can disagree with it. Flatten, then sign, in that order, because flattening after signing changes the bytes and breaks the signature.

Choosing what you actually need

Situation What is appropriate
Internal approval, delivery note, permission slip A signature image is fine
Freelance contract, NDA, rental agreement Image plus an emailed audit trail is the norm
Anything a regulator, court or bank will examine A digital signature from a recognised CA
A document you will send onward or merge Flatten first, then decide

Most of daily life sits in the first two rows. The mistake is not using a signature image, it is believing a signature image is doing something it is not, and skipping the audit trail because the document "is signed".

Doing it without handing the document over

The awkward part of signing online is that the documents involved are the sensitive ones by definition. Offer letters, tenancy agreements, medical consent forms and anything with a bank detail on it are exactly the files people drop into whichever free signing site is top of the results.

Placing a signature on a page is drawing an image at coordinates, and flattening is rewriting content streams. Neither needs a server. NextTool's PDF signing and flatten tools do both in the browser tab, so the contract stays on your machine.

What that cannot give you is the audit trail, because there is no third party observing. If you need a record that a specific person opened and signed at a specific time, you need a service that keeps one, and you are choosing to trade the document's privacy for that evidence. That is a reasonable trade when the evidence is the point. It is worth making it knowingly rather than by default.

Further reading

Do it hereSign PDFDraw, type or upload a signature and place it on any PDF pageOpen

Written for NextTool. Every statement about a tool is checked against that tool's code, and corrections are dated above. How guides are written

pdfsignaturesverification