Stronger assurance without collecting more.
Most identity systems buy confidence by holding more data about more people. The work described here starts from the opposite constraint: prove what has to be proved, disclose as little as the proof requires, and keep no record that is not needed to keep the promise made.
The public interest case, in one paragraph
A credential proves that somebody holds a credential. It does not prove that the person holding it is the person it was issued to, which is the question that actually matters at a login, a payment, an approval or a door. Closing that gap usually means more biometrics, more retained data and more central registers, and each of those is a new liability for the citizen and a new target for an attacker. The alternative is to make the check stronger at the moment it happens and weaker in what it leaves behind, and that is a design problem rather than a policy one. It is the design problem this work is about.
Areas of work
Digital identity
Confirming the authorised human at the moment of an action, weighing several independent signals so that one stolen factor is never enough, and scoring trust through a session rather than only at the door.
Authentication
Passwords prove knowledge and passkeys prove possession of a device. Neither proves a person is present. The distinction matters wherever a policy assumes strong authentication means a strong link to a human being.
Verification and provenance
Binding a physical or digital item to a live record and the rules attached to it, so that what a mark means can be changed, refused or withdrawn after it has been printed. A printed data-carrying code cannot do this, and a photograph of one is as good as the original.
Privacy engineering
Selective disclosure is the best idea in digital identity and it has a hole in it: answer enough narrow questions about a date of birth and the date of birth has been given away. Metering cumulative disclosure rather than policing each question is the subject of a pending application, GB2619018.1.
Data retention as a property, not a promise
Every service publishes a statement about what it does not keep, and that statement is a sentence somebody typed once. Deriving it from the running database, and refusing a schema that could hold the thing, makes the promise and the mechanism the same file. Five pending applications cover this.
Refusals that do not teach an attacker
A system that explains why it said no is a system that can be queried until it has explained how to pass. Deciding rule by rule how much of a reason can safely cross back is the subject of GB2619020.7 and GB2619062.9.
Age and entitlement assurance
Proving somebody is old enough happens once, at a moment, and says nothing about who holds the account an hour later. That gap is why checked accounts are quietly sold and shared, and is the subject of GB2619674.1.
Detecting forged marks without tracking anyone
A forged certification mark is not found by inspecting it. It is found when people in the field keep scanning something the register has never heard of, and that pattern is visible without following any individual. GB2619034.8.
Regimes this work sits inside
Named because they set the constraints the technology has to satisfy. Naming a regime is not a claim of certification, approval, endorsement or any relationship with the body that administers it, and nothing on this site should be read as one.
| Regime | Why it bears on this work |
|---|---|
| UK identity verification for company directors and filers | Millions of people must verify their identity to Companies House, and the mechanics of doing that at scale, for people who are not technologists, is a verification problem before it is a compliance one. Written up under Research and commentary. |
| The EU digital identity wallet | Selective disclosure moves from a good idea to a legal expectation, which makes the cumulative-disclosure problem above a live one rather than a theoretical one. |
| Online safety and age assurance | Highly effective age assurance has to be accurate, robust, reliable and fair at once, and a check that is accurate at a moment but transferable an hour later fails the second and third of those. |
| Data protection | Data minimisation is the principle that most identity architecture quietly fights. The retention work above is an attempt to make it structural rather than a policy document. |
What can be checked
Nothing on this page needs to be taken on trust. The sixty-one pending applications are listed with their numbers, applicants and register links at Patents. The companies that filed them are at Companies and ventures with their Companies House numbers. The reasoning behind several of them is written up at length under Research and commentary, to the published titles and general mechanism only. How each class of claim on this site is evidenced is set out at Evidence.
Contact for public sector enquiries
Departmental, regulatory and academic enquiries go to hello@louisjamesdavis.com. Briefings on any of the areas above are given without charge.
Institutional, investment and partnership enquiries.
Four kinds of enquiry reach Louis-James Davis at the same address. Say which one you are in the first line and it is answered faster.
- Investors What you invest in and the stage you come in at.
- Partners What you would want to build or integrate.
- Press Your outlet and your deadline in the first line.
- Working together What you have built. Working code beats a CV.