[enhancement]: Spam filter OCR and QR-code reading

[TaaviE]

Which feature or improvement would you like to request?

I'd love to see really simple/barebones OCR and QR link/text extraction built into Stalwart. This would make it somewhat harder to bypass the spam filter by just using images.

Is your feature request related to a problem?

Recent phishing waves have started using images to bypass the spam filter. While some other approaches might be able to mark letters like that suspicious, being able to actually see the content linked would be a great improvement.

Code of Conduct

• I agree to follow this project's Code of Conduct

[mdecimus]

Although implementing these enhancements shouldn't be too much work, my only concern is the latency these checks will add to message intake. But I'll check if there is a lightweight Rust library for OCR.

Regarding the QR spam, can you attach a few examples to this issue? I haven't seen this in the real world. Also, what is usually encoded in the QR, is it a link or something else?

[TaaviE]

Although implementing these enhancements shouldn't be too much work, my only concern is the latency these checks will add to message intake. But I'll check if there is a lightweight Rust library for OCR.

That's a fair point, though the trade-off is likely okay for some (especially smaller operators).

I suspect it might be a bit complex to implement some kind of lookup table and sampling system for more compute-intensive scans?

Regarding the QR spam, can you attach a few examples to this issue? I haven't seen this in the real world. Also, what is usually encoded in the QR, is it a link or something else?

Here's a recent example (fake tax return notice):