Yes, but not only.

We are used to relying on reputable sources online. However, for many legitimate reasons, these may delete, modify or move the various documents they are vouching for. Long term, we end up relying on third-party repositories of unknown trust, redistributing information.

We trust documents we find online simply because they look authentic or because we found them in a website that seems legitimate. We are unable to distinguish genuine documents stored by a honest third-party from convincing fakes with backdated metadata.

This is a weakness that sooner or later, will be exploited.

Because we are all late to the party.

The need for authentication in the archival landscape has long been known. The folks at LOCKSS drafted their threat model two decades ago, the LTANS working group concluded fifteen years ago, and Webrecorder integrated trusted timestamping several years ago.

However, most data online today is still not obviously authenticated. Digital archiving at scale is hard, budgets are small and issues like copyright are limits that stop content-oriented approaches. Today bad actors have all the capabilities to leverage these weaknesses.

We aim for a metadata-oriented approach that integrates on top of existing document repositories. We want anyone to be able to build today a tamper-evident index of their digital artifacts, for cheap, and in a way fit for long-term storage and large scale distribution.

We believe that the ideas underlying Certificate Transparency, Sigsum and Sigstore have a role to play in building tamper-evident historical records. They are tracking every year billions of records in transparency logs, a scale comparable to existing digital archives.

Web archiving today is built with WARC files together with CDX index files matching metadata with SHA-1 or MD5 checksums. There exists many more standards for digital archives, each pairing items to their metadata. We do not propose replacing these standards.

We propose to append supplemental index files, text-based, capturing just enough metadata together with cryptographic hashes, signatures and trusted timestamps. We want to empower anyone to build and share these indexes, which are structured as transparency logs.

We are deliberately trying to keep things simple then later add value.

We are not a root of trust.

We believe that distributed governance is key. We aim to use transparency logs to distribute trust between national institutions, non-profits, and private individuals.

We design to integrate with the existing transparency ecosystem. We believe that logs only need to monitor that new entries added are never backdated. We aim to provide recovery paths from partial log corruption, as well as several backup plans against catastrophic log failure.

This will require transparency-enforcing tools and clients. We want to enable anyone to write their own, to create software diversity. This will require open standards and we aim to assist the wider archiving community at making fast progress on these issues.

We want to enable both the small scale and the large scale by enabling clients to interact offline with slices of the index. We want to enable anyone to operate their own archive, large or small, and build tamper-evident logs of the artifacts they care about.

The output of Archive Transparency can be described as a large plaintext index of all the item hashes and metadata that people want to keep around, with a proof of past provenance.

This can be leveraged by third-parties to build content addressable storage as well as, at the cost of generating and storing document embeddings, efficient retrieval with full-text search.

We believe that this can open a path to search through digital archives at scale without interacting with the archive itself. This can both enable lower costs for digital archives, as well as great value for users and opportunities for others to build independent services.

The problems surrounding digital archives can be described as century-scale.

We do not aim to solve everything revolving around digital archives, but we do aim to bring into the long-term innovations like tamper-evident transparency logs.

Scaling file formats and protocols to several decades is not a trivial task as we cannot predict the future. This is in some fundamental way a guess at what will last, or not.

We still believe that some meaningful choices are possible to make, from using text files as our preferred way of storing metadata, extensible designs, forward-compatibility, etc.

Call it a ledger if you want!

Blockchain technologies spawned many innovations making full use of modern cryptography. However, we are solving for a different kind of problem. Yes, we are using append-only immutable trees of hashes. No, we are not building a blockchain-based system.

We believe here that our limiting factor is not good cryptography.

We are making opinionated choices specific to long-term digital preservation and believe that transparency logs are a better fit for this. We also believe that blockchain-based notary services will have some role to play, but we can not afford to rely on them.

We are still in the early stages of the project.

We are currently working on a research prototype to experiment with the design space available to us. The short term goal is to build a small-scale deployment of a million documents.

We invite you to contact us to help us scale to billions of records!