archive.rip

provable past provenance

Trust model

You are still early in this project.

This will be updated with a link once a formal model will be published.


We call archival notary a long-term digital notary fit for archival use.

Their target half-life can be 50 years or the century scale.

Why use an archival notary?

How do you know something is old when it is only a digital copy?

Publish now

Imagine you are in January and discover a document online that proves something important to you. It is hard to know if this document is authentic or not, you will investigate and have results in June.

However there is on-going world events that will affect your results as they happen in the next few months. They will likely affect your work.

You know that in June, people will argue that you have produced fake documents after the events unfolded to arrange your conclusions.

You decide to use a digital notary to get a cryptographic receipt that proves that this very real document existed in January.

You also decide to notarize your notes, keeping them private but publishing their digital integrity as you make them.

Argue later

When June arrives, you can show that the document existed and that you already wrote about them, before the world events unfolded.

This, while not proving the data you notarize are genuine, enables you to construct better arguments around provenance and authenticity.

Publishing integrity of your data is cheap for you to do today and offers many benefits to all future discussions relying on your data.

This effect only gets better the longer the cryptographic receipt stays available and more powerful if we pool our integrity together.


To build trust in past provenance you need people that will:

For step one, see the sibling of this page:

For step two, not all digital notaries are fit for long term use.

You need an archival notary.

What is an archival notary?

The archival notary must remember for you:

The important property is records produced are all public by default.

The record producer is separate from the record publisher and receipts include a proof of publication. This enable clients to enforce that records are discoverable without interacting with the publisher.


When you find a record in an archival notary, you learn with certainty:

You still need the data to make sense of a record and a fingerprint.

The archive, libraries, and your hard drive at home still need to exist.

Observers

You rely on the reputation of the observer (who recorded) to learn:

The observer can be ephemeral and access the data once to prepare the records that contains cryptographic digests of the data.

Mirrors

The publisher is transparent in the way that it maintains all records accessible to monitors that checks the reputation of these observers.

This is done by keeping a mirror copy of all records, in a tamper-evident, non-repudiable journal to make bad behavior discoverable.

You then rely on the reputation of the mirror (who published) to learn:

For an observer e-mail address, you rely on the publisher to check that an observer owns the claimed address, before accepting its records.

The mirror only store the records and is not trusted with time. It is its journal, a public ledger, that is trusted with keeping time.

We recommend using a transparency log to implement this journal: they are a simple model already adopted by the industry and the support of the WebPKI, the basis of the trust model of the Internet.

Monitors

The monitor role can be operated by third-party auditors or by the observers themselves, willing to verify the good reputation of the publisher by using the fact all records are public.

Given an insider threat, a private key being leaked or stolen, an observer could notice fake records published in its name. Similarly, a publisher that neglects its obligations or accept backdated records, can be noticed. This deters attackers by making abuse of trust discoverable.

This trust through transparency enables cryptography in the long term.


Verifying reputations is unavoidable as monitoring is always necessary.

This is true whenever digital trust integrates with the real world.

If you cannot get the full set of all the records an archival notary has ever published, attackers will sooner or later exploit this for abuse.

Hence the emphasis on publishing rather than anchoring: notaries can swap between different time-keeping methods or different digital identities with similar properties, while you always need a public shared state to prevent attackers from hiding their bad behavior.

Scaling in time

This story around transparent records of "what, when, who and how" enables meaningful conversations around archival notarization.

This enables different narratations of a shared history in a world where we know future bad actors will attempt to fake the past at scale.

The records being both public, authenticated, and in a resilient trust model, is what will enable an archival notary that scales in time.

If we agree that cryptographic digests are here to stay, future mirrors and observers can coordinate to anticipate change and keep up with the world evolving, without invalidating all previous records.

As long as someone human is here to keep the lights on.

Do I need to remember this to use it?

No, this is mostly informative.

Think about web browsers: they do a lot of work that enables you to, when you visit your online bank, you know that your computer is talking to other computers your bank control.

As a end-user, you rely on the security of the tools you use. You rely on the vendor reputation of the people building the software.

You trust green "verified" checkmarks if you can believe that experts of the field or the wider society have checked them for you.

As we, citizens, delegate our understanding of these niche topics impacting our lives, we need to verify that technical choices are truly made in our collective interest and that we control what we rely on.

Now and for our future.

Mechanized intelligence is making us doubt the digital world. We will need more green "verified" checkmarks build by someone else's software.

This page is here to help you ask questions about these checkmarks.

The rest is about being careful in who you trust when you choose software and question who is in control of the machine you rely on.


Think of archive.rip as one voice on the topic of archival notaries.

We advocate for verifiable trust through transparency to make cryptography work for the people and not the other way around.

However we together should cultivate more than one solution to this.

If you are working to make provable past provenance more accessible to the public, please contact us to be listed as an option below.

See projects that we believe pursue goals similar to us:

We also cultivate our own software link and want you to do the same.

Why only rely on technology?

This would be a hard bet.

When examining how digital machines changed us in the past 50 years, how can we predict the next 50 years?

Under the current promises of intense technological change pushed by the idea of mechanical intelligence, our own lifetimes as human individuals, will likely be longer than today's idea of technology.

While not certain, other forces can push for chaos in the following decades and make us uncertain of anything digital in the world.

Human memory held by us individuals is as much as the larger end goal than a powerful tool that we can use at our advantage.

This can be made real through art, education, or physical artifacts that we as humans could get attached to, value and collect.

Please reach out if you have ideas and want to contribute, or order a printed book link of cryptographic proofs to support this project.

What about other kinds of notaries?

There are other kinds of digital notaries and applications for them.

This section details how other technologies, such as timestamping services and blockchain-based finance, relate with our nomenclature.


click here to expand this section

A/ Opaque & Anonymous Notaries

Time Stamping Authorities, rfc3161's TSAs are time-keeping services that when shown a message, can return the same message, signed by the authority with its own local time. This is a stateless process.

TSAs are build to be truthful time keepers, often leveraging costly hardware, but the cryptographic receipt they delivers ignore "who recorded?" and in that aspect, they can be called anonymous.

The receipts produced by the authority are private to their submitter and can not be discovered by a third-party. When shown any receipt, one can verify it has been seen by the authority but nothing more.

They can be called opaque in that way.

Remembering "who recorded?" and making this property discoverable is important to make an archival notary: for example, when two observers look at the same collection, but disagree on the exact content, it is important to remember that two different parties did the same work from their point of view, but obtained different results.

Choosing to "remember who" informs future researcher inspecting records of the different point of views and concurrent narrations of our present day artifacts, just like historians compare different sources.

This makes a notary helpful to build attestation of historical records.


However this goes beyond the original intent of time-keeping services.

Time-keeping services are useful to build attest time of existence.

They can be used as simple opaque anonymous notaries. For example, you can get a cryptographic receipt of a message together with a trusted time, signed by the federal government of Belgium link.

There exist blockchain-based time-keeping services build on the same opaque and anonymous model. Their value proposition is to substitute well-known recognizable institutions with financial cryptographic markets using more electricity than all of Belgium.

Beyond this usual objection, most of the blockchain-based systems benefits (notably their resistance to an authority being corrupt) can be obtained by enforcing transparency of the cryptographic receipts in a public ledger, enabling monitors to verify the authority good behavior, which is the model we are advocating for.

There are also other arguments that questions the use of incentive-based networks for archiving. It is unclear how these systems will age in time or react to large-scale change, as incentives shift and society evolves.

The complexity of financial cryptography can make it unfit in many contexts. There are also concerns in the concentration of power to financial markets, themselves prone to rapid change and instabilities.

Finally, these are only time-keeping services: once records have been collected and logged in public journals, the hard part have been done.

The time-keeping services should be diverse and "anchoring" should be a component that can be replaced by other technologies without harm.

The core should be about trusted publishing; not time-keeping.

B/ Transparent & Authenticated Notaries

Most of the limits highlighted above of the opaque and anonymous notaries, disappear if they become transparent and authenticated.

An authenticated notary accepts only signed submissions and refuses to notarize a message not attached to a digital identity. This identity can be pseudonymous (submitter identified by their public material) but such authenticated notary can also promise to validate a label, checking that the submitter owns a domain name or another form of identity.

However using digital identities gives opportunities for abuse: an insider to the submitter organisation notarize harmful forgeries, the notary itself emulate "submitter puppets" faking other people's identities, neglect lets an attacker steal an identity, etc.

To mitigate these issues, the notary builds a public ledger of the records it has notarized, that is tamper-evident and non-repudiable, and records that can not be proved as included in the notary's ledger are considered as invalid. This ledger (the journal or log in our model) enables third-party audits of the records being produced: submitters can check what are the records are produced under their name, monitors can check for bad behavior from the notary, etc.

This makes a transparent notary.

This also solve many of the frustrations of pure time-keeping services, as it enables better stories around key rotations, recovery from a bad notary, or discovery of other existing records produced by other people.

This model we are advocating for here, is inspired by prior art in:

This also put us in a close lineage to previous art in linked timestamping including ArchiSig; the Long-Term Archive and Notary Service LTANS working group, which inspired modern eIDAS qualified electronic archiving systems for legal compliance like the BSI TR 03125 TR-ESOR.

C/ What about privacy?

Transparent notaries publish all their output and are not private.

This has benefits against misuse of the cryptographic materials by protocol participants, but makes de-facto all artifacts public. In a way, just like an attacker cannot hide harm and abuse, participants are forced to work under the assumption that the digests and the metadata they submit, will be public and accessible by anyone.

The data observed by submitters can stay private, only the cryptographic digests of the data are published as well as the selected metadata. This makes sense as a default for use cases that involve public collections.

However for use cases where the submitter is under a high level of threat, the submitter must be aware that a bad actor that already have the same data, can notice the submitter publishing the digest of that data. For example, a monitor can check if the digest of a specific picture is seen by the notary or not. This is by design.

For privacy, it is recommended that an observer under such threat by bad actors bundle first its data in a compressed archive (in a .zip format, for example) with at least one secret file (a picture taken with a phone at home) inside, then submit the archive cryptographic digest.

Future integrations could be build into tooling and software to automate a solution to this privacy concern. Building a privacy-preserving technology is on topic long term, but for this model use case, unconditional publication is simpler to implement.

This project is still a work in progress:

  • please contact us link to contribute and improve it!

Return to the home page: archive.rip