eIDAS: Building Trust or Invading Privacy?
epicenter.works
We analysed the new technical framework for the European Digital Identity Wallet, revealing severe shortcomings that threaten user privacy and contradict the regulation's intent.

cross-posted from: https://feddit.org/post/317047

in February 2024, the EU Parliament adopted the eIDAS regulation, creating the framework for a “European Digital Identity Wallet”. This digital Wallet will enable citizens to identify themselves in a legally binding manner, both online and offline, sign documents, login into websites and share personal data about them with others. Recently, the European Commission published the Architectural Reference Framework (ARF) 1.4 for the technical implementation of the Wallet.

The success of the EU Digital Identity Wallet depends on its ability to gain citizens’ trust and establish a resilient infrastructure in our current data-driven economy.

“However, after our analysis, we believe that this goal has been missed,” says the digital rights group Epicenter Works.

“We see severe shortcomings in the ARF that either contradict the regulation or ignore important elements of it. These issues, if left unaddressed, could significantly undermine user rights and privacy.”

  • huginn@feddit.it
    8·
    5 months ago

    Most of these make sense and are definitely blockers for this ever releasing but -

    Remove the concept of the Pseudonym Provider and ensure pseudonyms are generated and stored locally without the possibility of linking back to real identities.

    Correct me if I’m wrong but this data all has to be signed somewhere right? Like the eID contains cryptographically signed assertions about the user in some standard (JWT?) format.

    What use is signing the assertions locally? There would be no way to tell if the citizen actually had any valid id at all. A pseudonym provider is the privacy layer that allows for signing of new tokens after ensuring the validity of the old.

    How could you sign an anonymous token using a valid one without it being linked back to the valid one? It seems like impossible constraints.

    Am I totally off base here?

    • originalfrozenbanana@lemm.ee
      82·
      5 months ago

      No you’re right. The ARF just ignored that constraint and intentionally built in a back door here. From the linked article:

      However, the current ARF stipulates that law enforcement authorities can retroactively trace pseudonyms back to their legal identity. The provisions therefore „strongly contradicts the legal requirements,“ epicenter.works writes.

      • huginn@feddit.it
        4·
        5 months ago

        Agreed that law enforcement should not be involved but the quote I posted was also from the article and it seems impossible.

          • huginn@feddit.it
            2·
            5 months ago

            It’s impossible to do without exposing a private signing cert to everyone, yes. That’s the issue.

            You can’t do asymmetric key signing anonymously and with a central issuer.

            So either you have to just trust the assertions (0 security) or you have to have a trusted issuer (not anonymous)

            A pseudonym issuer is a trusted issuer. There’s no way to do it otherwise. You have to trust someone to make this kind of system work.