Adding back TripleDES decryption

[pc-coholic]

Background: If you have an NDA with NXP and they provide you with controlled documents, those documents will have a big stamp applied all over them, as to be able to track them. But: they also apply "certificate security" to the documents, requiring the user to have their client certificate(s) installed and using a PDF-viewer that can decode such PDF files for your viewing pleasure.

Unfortunately, NXPs official recommendation is: Either use Windows and the official Adobe Acrobat (Reader) or - if you absolutely need to use Linux - there is this old-ass 9.5.5 version for unix systems. Good luck.

pyHanko seems to be a new, third option to this problem...

• Use pyHanko 0.25.3
• Fix the detection of --force-flag (see CLI: Parse --force flag #537)
• Use the git-version of oscrypto that fixes the detection of OpenSSL >= 3.0.10

With this setup, pyhanko decrypt pkcs12 --force will decrypt the PDFs provided by NXP.

Going this route is - unfortunately - required, since pyHanko 0.26.0 dropped support for oscrypto and by extension DES, TripeDES and RC2. NXP is using 3DES.

Long story short: Is the decision on dropping 3DES final?

It is my understanding, that dropping the above 3 algos was mostly pinned on "old and outdated" but also "need to get rid of oscrypto"...

Just for fun, I tried plugging in a 3DS decryptor using the "dangerzone" (cryptography.hazmat.decrepit.ciphers.algorithms):

def tripledes_decrypt(key, data, iv):
    cipher = Cipher(TripleDES(key), mode=modes.CBC(iv))
    decryptor = cipher.decryptor()
    return decryptor.update(data) + decryptor.finalize()

This will throw a RuntimeError: generator didn't stop - but between me not knowing what I am doing and NXP doing weird things with their files, I don't really know where to look for this...

Sooo... Any chance we could see a return of 3DES? Pretty, pretty please? :)

[MatthiasValvekens]

Ah, someone using the public-key based document encryption feature in the wild? Interesting... 😉

I see your point. Adding back the dependency on oscrypto is something I'm not keen on doing, given the maintenance status of that project. That said, when I first integrated support for the document encryption feature, 3DES was not exposed through pyca/cryptography (hence the dependency on oscrypto for that), so reinstating it through pyca/cryptography is something I'd be willing to consider. Getting a test document from a reliable source could be slightly annoying, though.

I'll also take a look at #537.

[MatthiasValvekens]

Yeah, chances are I'll end up having to generate my own test files, which I can do, but then that also means that I'd need to find another implementation to test those files with, so we're back at square 1.

That said, there's a decent chance that the algorithm support in something like iText is mostly dictated by what's available in the JVM you run it on. Pretty sure they don't have tests for 3DES, but I have my reasons for suspecting that this'll probably "just work" for decryption purposes. Can't really check that right now, though.

[MatthiasValvekens]

@pc-coholic I haven't tested this at all, but can you give the code on branch feature/reinstate-old-algos a spin?

[pc-coholic]

The decryption process works like a charm for me :-)

Thank you for adding it back in so quickly!

[MatthiasValvekens]

Alright, good news!

In the meantime, I also managed to generate some test docs. Still need to do some compat testing on those, but I can probably turn all of this around pretty soon.

[MatthiasValvekens]

This is in 0.27.1 now.