Hello,
I would love to find out how to do pseudonymization on my Make… scenarios.
But I can’t find anything. Quite a lot about anonymization but it’s irreversible so not interesting for me.
Have any of you already tried? succeeded? With which module or api?
I’d like to be able to identify personal or sensitive data in a text, encapsulate it with pseudonymization, process the document or query with pseudonymized data, and then put it back in the answer.
For example:
Processing an invoice :
An invoice arrives on Make,
I extract the information via OCR
I transform the personal information: John Smith becomes Benoit Zig, his phone was 0320203030 and he becomes 09488484848, etc.
The module retains the decryption key so that the correct data can be retrieved at the end.
I perform actions on the anonymized document: e.g. send to Google docs.
then, at the end, I put the correct data back into the document so that I can file it.
If you don’t find a dedicated solution, perhaps give the problem to an AI/LLM/chatgpt “You’re the Chief Security Officer and you are excellent at converting phone numbers, names, passwords into alternate fake values.”
Yes but a lot of my clients don’t trust Azure, OpenAi, Google… about data privacy.
(GDPR, Patrioct Act, …)
That’s why i’m searching for european actor or French Actor.
i’m stuck.
I think I’d trust MS Azure more than I’d trust Make.com with my data; based on the formers ability to ensure isolation and the very public consequences of getting it wrong.
But yes it is tricky to help clients/employers work through this topic for all their systems.
Encryption and Pseudonymization are two different things and has different use-cases.
Encryption?
Encryption is like locking a box with a key. Only someone with the correct key (the encryption key) can unlock the box (decrypt the data) and see what’s inside.
Encryption is reversible provided you store the encryption key to use in a decryption process.
Encryption doesn’t allow you to view the rest of the file without decrypting it first.
Example: Encrypting a credit card number before storing it in a database prevents unauthorized access.
Pseudonymization?
Pseudonymization is like replacing names in a story with made-up names. You can still read the story without knowing the real names, but you can’t go back and change the made-up names to the real ones unless you keep a list of replacements.
Pseudonymization is not reversible, unless you store the position in the document and the original text that was replaced somewhere, so that you can replace it back later.
Pseudonymization, without encryption, allows you to view the rest of the file because it is still in the same format.
Example: Replacing patient names with numbers in a medical database to protect privacy while still being able to analyze the data.
You are confusing two different things and confusing us
You are confusing us because you are not clear on what you want, either or both, and how you want it implemented.
Encryption is often the primary choice for securing sensitive data
Pseudonymization can be a measure when you need to share the rest of the data while maintaining privacy. Pseudonymization is generally not reversible.
Hope this helps! Let me know if there are any further questions or issues. Note: I see hundreds of posts, notifications, and messages daily on this forum, so if I missed your reply, please message me to look at your reply.
P.S.: Did you know, the concepts of about 70% of questions asked on this forum are already covered in the Make Academy. Investing some effort into it will save you lots of time and frustration using Make later!
I know the difference very well and I only talk about PSEUDONIMYSATION in my need.
Because I need it to be reversible and still readable during the process.
Thanks for trying @LinkYourTech but i don’t need encryption
I gave you an example use case at the beginning.
But I think that in reality it seems impossible via Make.com.
As you said @drnic , I’ll probably have to manage it first. Which doesn’t suit me at all, because it makes the whole thing much more complex…
If anyone out there has any ideas, I’m really looking for them.
I’d even be willing to pay ^^
Currently facing the same issue. I’ve tried to use RegEx patterns with the Match Patterns Module, but my input data is not enough structured or doesn’t follow enough rules to fit in this solutions.
I now reached out to a Berlin-based company called Langdock as they host different LLMs in Europe with ISO27001, GDPR compliant etc and consider to just call their API for pseudonymization. As your clients seems to be very sensitive regarding privacy, that still might be not enough, but you can check it out. One of my clients is already experimenting with a local LLM as he wants to use it for other purposes. So I might use the local one if Langdock is not compliant enough for my clients regulations.
If you’ve found a proper solution in the meantime I would be very interested!