Validation functionality

Note

Before reading this, you may want to take a look at Factors in play when validating a signature for some background on the validation process.

Danger

In addition to the caveats outlined in Validating PDF signatures, you should be aware that the validation API is still very much in flux, and likely to change by the time pyHanko reaches its beta stage.

General API design

PyHanko’s validation functionality resides in the validation module. Its most important components are

While you probably won’t need to interface with DocumentSecurityStore directly, knowing a little about EmbeddedPdfSignature and SignatureStatus is useful.

Accessing signatures in a document

There is a convenience property on PdfFileReader, aptly named embedded_signatures. This property produces an array of EmbeddedPdfSignature objects, in the order that they were applied to the document. The result is cached on the reader object.

These objects can be used to inspect the signature manually, if necessary, but they are mainly intended to be used as input for validate_pdf_signature() and validate_pdf_ltv_signature().

Validating a PDF signature

All validation in pyHanko is done with respect to a certain validation context (an object of type pyhanko_certvalidator.ValidationContext). This object tells pyHanko what the trusted certificates are, and transparently provides mechanisms to request and keep track of revocation data. For LTV validation purposes, a ValidationContext can also specify a point in time at which the validation should be carried out.

Warning

PyHanko currently uses a forked version of the certvalidator library, registered as pyhanko-certvalidator on PyPI. The changes in the forked version are minor, and the API is intended to be backwards-compatible with the “mainline” version.

The principal purpose of the ValidationContext is to let the user explicitly specify their own trust settings. However, it may be necessary to juggle several different validation contexts over the course of a validation operation. For example, when performing LTV validation, pyHanko will first validate the signature’s timestamp against the user-specified validation context, and then build a new validation context relative to the signing time specified in the timestamp.

Here’s a simple example to illustrate the process of validating a PDF signature w.r.t. a specific trust root.

from pyhanko.sign.general import load_cert_from_pemder
from pyhanko_certvalidator import ValidationContext
from pyhanko.pdf_utils.reader import PdfFileReader
from pyhanko.sign.validation import validate_pdf_signature

root_cert = load_cert_from_pemder('path/to/certfile')
vc = ValidationContext(trust_roots=[root_cert])

with open('document.pdf', 'rb') as doc:
    r = PdfFileReader(doc)
    sig = r.embedded_signatures[0]
    status = validate_pdf_signature(sig, vc)
    print(status.pretty_print_details())

Long-term verifiability checking

As explained here and here in the CLI documentation, making sure that PDF signatures remain verifiable over long time scales requires special care. Signatures that have this property are often called “LTV enabled”, where LTV is short for long-term verifiable.

To verify a LTV-enabled signature, you should use validate_pdf_ltv_signature() instead of validate_pdf_signature(). The API is essentially the same, but validate_pdf_ltv_signature() takes a required validation_type parameter. The validation_type is an instance of the enum pyhanko.sign.validation.RevocationInfoValidationType that tells pyHanko where to find and how to process the revocation data for the signature(s) involved1. See the documentation for pyhanko.sign.validation.RevocationInfoValidationType for more information on the available profiles.

In the initial ValidationContext passed to validate_pdf_ltv_signature() via bootstrap_validation_context, you typically want to leave moment unset (i.e. verify the signature at the current time).

This is the validation context that will be used to establish the time of signing. When this step is done, pyHanko will construct a new validation context pointed towards that point in time. You can specify keyword arguments to the ValidationContext constructor using the validation_context_kwargs parameter of validate_pdf_ltv_signature(). In typical situations, you can leave the bootstrap_validation_context parameter off entirely, and let pyHanko construct an initial validation context using validation_context_kwargs as input.

The PAdES B-LTA validation example below should clarify that.

from pyhanko.sign.general import load_cert_from_pemder
from pyhanko.pdf_utils.reader import PdfFileReader
from pyhanko.sign.validation import (
    validate_pdf_ltv_signature, RevocationInfoValidationType
)

root_cert = load_cert_from_pemder('path/to/certfile')

with open('document.pdf', 'rb') as doc:
    r = PdfFileReader(doc)
    sig = r.embedded_signatures[0]
    status = validate_pdf_ltv_signature(
        sig, RevocationInfoValidationType.PADES_LTA,
        validation_context_kwargs={'trust_roots': [root_cert]}
    )
    print(status.pretty_print_details())

Notice how, rather than passing a ValidationContext object directly, the example code only supplies validation_context_kwargs. These keyword arguments will be used both to construct an initial validation context (at the current time), and to construct any subsequent validation contexts for point-of-time validation once the signing time is known.

In the example, the validation_context_kwargs parameter ensures that all validation will happen w.r.t. one specific trust root.

If all this sounds confusing, that’s because it is. You may want to take a look at the source of validate_pdf_ltv_signature() and its tests, and/or play around a little.

Warning

Even outside the LTV context, pyHanko always distinguishes between validation of the signing time and validation of the signature itself. In fact, validate_pdf_signature() reports both (see the docs for timestamp_validity).

However, since the LTV adjudication process is entirely moot without a trusted record of the signing time, validate_pdf_ltv_signature() will raise a SignatureValidationError if the timestamp token (or timestamp chain) fails to validate. Otherwise, validate_pdf_ltv_signature() returns a PdfSignatureStatus as usual.

Incremental update analysis

Changed in version 0.2.0: The initial ad-hoc approach was replaced by a more extensible and maintainable rule-based validation system. See pyhanko.sign.diff_analysis.

As explained in the CLI documentation, the PDF standard has provisions that allow files to be updated by appending so-called “incremental updates”. This also works for signed documents, since appending data does not destroy the cryptographic integrity of the signed data.

That being said, since incremental updates can change essentially any aspect of the resulting document, validators need to be careful to evaluate whether these updates were added for a legitimate reason. Examples of such legitimate reasons could include the following:

  • adding a second signature,

  • adding comments,

  • filling in (part of) a form,

  • updating document metadata,

  • performing cryptographic “bookkeeping work” such as appending fresh document timestamps and/or revocation information to ensure the long-term verifiability of a signature.

Not all of these reasons are necessarily always valid: the signer can tell the validator which modifications they allow to go ahead without invalidating their signature. This can either be done through the “DocMDP” setting (see MDPPerm), or for form fields, more granularly using FieldMDP settings (see FieldMDPSpec).

That being said, the standard does not specify a concrete procedure for validating any of this. PyHanko takes a reject-by-default approach: the difference analysis tool uses rules to compare document revisions, and judge which object updating operations are legitimate (at a given MDPPerm level). Any modifications for which there is no justification invalidate the signature.

The default diff policy is defined in DEFAULT_DIFF_POLICY, but you can define your own, either by implementing your own subclass of DiffPolicy, or by defining your own rules and passing those to an instance of StandardDiffPolicy. StandardDiffPolicy takes care of some boilerplate for you, and is the mechanism backing DEFAULT_DIFF_POLICY. Explaining precisely how to implement custom diff rules is beyond the scope of this guide, but you can take a look at the source of the diff_analysis module for more information.

To actually use a custom diff policy, you can proceed as follows.

from pyhanko.sign.general import load_cert_from_pemder
from pyhanko_certvalidator import ValidationContext
from pyhanko.pdf_utils.reader import PdfFileReader
from pyhanko.sign.validation import validate_pdf_signature

from my_awesome_module import CustomDiffPolicy

root_cert = load_cert_from_pemder('path/to/certfile')
vc = ValidationContext(trust_roots=[root_cert])

with open('document.pdf', 'rb') as doc:
    r = PdfFileReader(doc)
    sig = r.embedded_signatures[0]
    status = validate_pdf_signature(sig, vc, diff_policy=CustomDiffPolicy())
    print(status.pretty_print_details())

The modification_level and docmdp_ok attributes on PdfSignatureStatus will tell you to what degree the signed file has been modified after signing (according to the diff policy used).

Warning

The most lenient MDP level, ANNOTATE, is currently not supported by the default diff policy.

Danger

Due to the lack of standardisation when it comes to signature validation, correctly adjudicating incremental updates is inherently somewhat risky and ill-defined, so until pyHanko matures, you probably shouldn’t rely on its judgments too heavily.

Should you run into unexpected results, by all means file an issue. All information helps!

If necessary, you can opt to turn off difference analysis altogether. This is sometimes a very reasonable thing to do, e.g. in the following cases:

  • you don’t trust pyHanko to correctly evaluate the changes;

  • the (sometimes rather large) performance cost of doing the diff analysis is not worth the benefits;

  • you need validate only one signature, after which the document shouldn’t change at all.

In these cases, you might want to rely on the coverage property of PdfSignatureStatus instead. This property describes the degree to which a given signature covers a file, and is much cheaper/easier to compute.

Anyhow, to disable diff analysis completely, it suffices to pass the skip_diff parameter to validate_pdf_signature().

from pyhanko.sign.general import load_cert_from_pemder
from pyhanko_certvalidator import ValidationContext
from pyhanko.pdf_utils.reader import PdfFileReader
from pyhanko.sign.validation import validate_pdf_signature

root_cert = load_cert_from_pemder('path/to/certfile')
vc = ValidationContext(trust_roots=[root_cert])

with open('document.pdf', 'rb') as doc:
    r = PdfFileReader(doc)
    sig = r.embedded_signatures[0]
    status = validate_pdf_signature(sig, vc, skip_diff=True)
    print(status.pretty_print_details())

Probing different aspects of the validity of a signature

The PdfSignatureStatus objects returned by validate_pdf_signature() and validate_pdf_ltv_signature() provide a fairly granular account of the validity of the signature.

You can print a human-readable validity report by calling pretty_print_details(), and if all you’re interested in is a yes/no judgment, use the the bottom_line property.

Should you ever need to know more, a PdfSignatureStatus object also includes information on things like

  • the certificates making up the chain of trust,

  • the validity of the embedded timestamp token (if present),

  • the invasiveness of incremental updates applied after signing,

  • seed value constraint compliance.

For more information, take a look at PdfSignatureStatus in the API reference.

Footnotes

1

Currently, pyHanko can’t figure out by itself which LTV strategy is being used, so the caller has to specify it explicitly.