De-identification
Your company licenses a copy of its entire operating record: mail, chat, files, tickets, and ledgers. Mikra transforms all of it so that identities change and the work stays intact. One real person becomes one stand-in identity in every source. You review a sample before any customer receives anything. The risk is lower, not zero.
- Email · Mar 14, 2024
- Before: From: Dana Whitfield <dana.whitfield@harborline-cpa.com>. Sending the bank reconciliation to Kessler Dental today.
- After: From: Priya Raman <priya.raman@alderbrook.example>. Sending the bank reconciliation to Marlow Dental today.
- Chat · #close-team
- Before: @dana.w Kessler Dental says invoice 4471 was paid twice.
- After: @priya.r Marlow Dental says invoice 4471 was paid twice.
- Ticket · AP-2291
- Before: Assignee: Dana Whitfield. Customer: Kessler Dental. Call (617) 555-0148.
- After: Assignee: Priya Raman. Customer: Marlow Dental. Call (000) 555-0199.
Black bars break the record.
Business records are valuable because they hold long workflows and relationships: who asked, who approved, who was late. They also hold personal information. Blacking it out lowers the risk, but it strips context and breaks the links between records. So the job is not finding names. It is transforming a whole body of records, with three goals.
- 01
Identity protection
Anything that can identify a person or a company is transformed, whether it does so directly, like a name, or through context, like a job title in a small team.
- 02
Meaning preserved
What is not sensitive stays unchanged, and every replacement fits its type and its context.
- 03
Consistency across documents
One real person or company becomes one stand-in identity everywhere, and two different people never merge into one.
Five methods, compared.
There are five common ways to handle personal information in records. Each trades privacy against how useful the result is and whether it still holds together across documents.
| Approach | Privacy | Usefulness of what is left | Consistency across records |
|---|---|---|---|
| Removal or blacking outThe value is deleted or covered. | High | Low | Low |
| Pattern maskingAnything shaped like an email or a number is replaced with a fixed mask. | Medium | Low | Low |
| Entity replacement, field by fieldEach detected value is swapped on its own. | Medium to high | Medium | Medium |
| Fresh synthetic dataNew records are generated to resemble the originals. | High | Medium | Low |
| Identity-preserving transformationEach person or company gets one stand-in identity across the whole record.Mikra’s approach | High | High | High |
Only the last approach protects identities and keeps the record usable as a whole. The ratings describe each approach done well. They are not measurements of any one system.
Seven stages, in order.
This is the process Mikra runs on a company’s records. Machines do the first pass. People check the hard cases and a sample of the result. You check last.
- 01
Inventory and segmentation
Every file is listed with its format and what it takes to rebuild it. Large files are split into parts. The inventory is the count everything is measured against later, so a file that fails cannot quietly drop out.
- 02
Detection and entity resolution
Detectors flag candidate identifiers across the whole body of records, not one file at a time. A second sweep links aliases and indirect mentions to the same person or company. That linking is called entity resolution. Finding a value and deciding whose it is are separate steps, tied together by one shared identity record.
- 03
Review of hard cases
Ambiguous candidates are confirmed, rejected, or relabeled. Aliases are merged. False merges are split. Degraded values, such as typos, look-alike characters, and odd spacing, are traced back to their owner.
- 04
Identity mapping
Each protected person or company gets one stand-in identity: a name, an email address, a username, and an employer that fit together. It is not a set of random field-by-field swaps. The mapping is fixed for the whole run.
- 05
Reconstruction and validation
Files are reassembled and checked against the inventory: rows and columns, sheet names, page counts, archive members. A file that opens is not proof that the identity inside it is right. So structure checks and meaning checks are separate.
- 06
Human review
Reviewers read a sample drawn from the inputs, not from the successes, so failed or missing files stay in the measurement. They mark exposed identifiers, unnecessary changes, wrong labels, fragmented replacements, two people merged into one, and anything that reads wrong.
- 07
Your review
Before the first delivery to any customer, you get a representative sample of the transformed data and a short description of the methods. You have 10 business days to point out anything we missed. We fix it before anything is delivered.
Raw data stops at Mikra.
Three parties, in order. Your company and Mikra hold the raw records. Customers receive only the de-identified data and what is built from it.
- 01
Your company
Keeps its records, its systems, and ownership. It sends a copy through Mikra’s secure upload. The records listed under “never uploaded” below stay out of it.
- 02
Mikra
Receives the raw records by encrypted file transfer, on storage Mikra operates itself: a Ceph storage cluster on Mikra’s own hardware in a colocation data center. Only Mikra staff have access to that storage, and access is logged. De-identification is run from that storage. When the de-identification models need more computing power, Mikra rents GPU hosts only from providers that hold a SOC 2 report. The raw records are seen only by Mikra staff and service providers who need them to review and de-identify, under written confidentiality. We delete the raw upload within 120 days after upload and keep the de-identified data and the products built from it.
- 03
Customers
Receive de-identified data and the products built from it. They never receive the raw upload, and their contracts bar them from re-identifying anyone.
Five questions, answered separately.
A strong score on one question cannot hide a failure on another, so the five are reported separately. Mikra will report them to your company with its sample. We have not yet published measurements of our own, so no score appears here.
- 01
Privacy
How much real identifying information is still exposed, weighted by harm. A missed government ID or bank account counts for more than a missed date.
- 02
Usefulness
How much was changed that should have been left alone.
- 03
File integrity
Whether each delivered file is still an intact, working version of the original.
- 04
Coverage
Whether every expected input was processed. Dropping the hard files cannot raise the score.
- 05
Coherence
Whether the result reads naturally and keeps each identity linked across documents, without merging two people into one.
Detection alone falls short.
These charts are about the problem, not about Mikra. They show how much a first pass misses, and how much entity resolution and review recover.
Source of every figure in this sectionPublished third-party evaluation, 2026. None of these systems is Mikra’s. Mikra has not yet published its own measurements.
Off-the-shelf detection still misses identifiers.
Seven systems ran on a public test set of 5,198 workplace messages holding 8,805 hand-labeled identifiers of five types: given name, family name, email address, username, and organization. Two measures are shown. Recall is the share of known identifiers a system found. Precision is the share of its flags that were correct.
| System | Identifiers found (recall) | Flags that were correct (precision) |
|---|---|---|
| Purpose-built transformation pipelineBest in the evaluation | 95.1% | 97.0% |
| Commercial de-identification tool | 92.6% | 96.5% |
| Open-source rule-based detector | 89.0% | 87.3% |
| Open-source entity-recognition model | 88.6% | 85.2% |
| Large general-purpose AI model | 88.2% | 95.4% |
| Mid-size general-purpose AI model | 85.0% | 97.0% |
| Small general-purpose AI model | 79.9% | 97.6% |
Messy enterprise data is harder.
A synthetic company dataset was built to be hostile: 600 employees and 420 customers recurring across HR, finance, technology, logs, and customer tables. Identifiers were hidden behind typos, look-alike characters, spelled-out digits, odd separators, and zero-width characters, which take up no space on screen. Decoys were planted, such as part numbers shaped like Social Security numbers and version numbers shaped like IP addresses. A widely used open-source detection tool ran in its default configuration, then again with entity resolution across the whole dataset and a review step.
| Measure | Default configuration | With entity resolution and review |
|---|---|---|
| Planted identifiers foundHigher is better | 74.3% | 89.2% |
| Flags that were correctHigher is better | 62.4% | 77.5% |
| Look-alike decoys wrongly transformedLower is better | 72.3% | 40.0% |
Some cases cannot be settled.
Real data wears down. A title such as “Ms.” can turn up as “MS.”, “M.s”, or just “M.”. So can “Mx.”. Once both have collapsed to “M.”, the information that told them apart is gone. Link the two records and you may merge two people into one. Keep them apart and you may split one person in two. No amount of care recovers what the data no longer holds.
- Ms.
- MS. · M.s · M.
- Mx.
- M.
Raw data stops before Mikra.
There are two tracks. On the standard track above, Mikra de-identifies. For a healthcare company’s patient or clinical records and a law firm’s client files, an independent provider de-identifies before Mikra receives anything. This track is not open yet.
- 01
Your company
A healthcare company’s patient or clinical records and a law firm’s client files do not go through Mikra’s upload. Your company sends them to an independent third-party de-identification provider. A law firm stays responsible for its duties to its clients and confirms what it may share.
- 02
Independent provider
De-identifies the records before Mikra receives anything. For healthcare, the provider signs a Business Associate Agreement with your company. Raw data stops here. The provider is not yet contracted, so this track is not open yet. Mikra accepts no records on it until a provider is contracted.
- 03
Mikra
Receives only the de-identified records. On this track Mikra never receives protected health information. The track runs under a Healthcare and Legal Addendum to the Data License Agreement.
- 04
Customers
Receive de-identified data and the products built from it, under contracts that bar them from re-identifying anyone.
The work stays whole.
A model team trains on how work moves from request to approval. Strip the structure and nothing is left to learn.
- 01
Threads
Every reply stays in its thread, and every thread stays in its channel.
- 02
Timestamps
The order of events and the time between them stay intact.
- 03
File and ticket links
A message that points to a file or a ticket still points to it.
- 04
Ledger relationships
Invoices, payments, and journal entries still reconcile against each other.
Some things never leave.
These records are never uploaded to Mikra. Patient or clinical records and a law firm’s client files move only through the healthcare and legal track.
- Patient or health records, outside the healthcare and legal track
- Biometric identifiers
- Card numbers and bank account numbers
- Social Security numbers and other government ID numbers
- Passwords and credentials
- Personal information of anyone under 18
- Export-controlled or classified information
- Attorney-client privileged material, outside the healthcare and legal track
- Files held for a client under a contract that forbids sharing
The risk is lower, not zero.
It is lower for three reasons. What identifies a person or your company is transformed: names, company and client names, contact details, account numbers, and similar identifiers. You review a sample before anything is delivered. Customers are contractually barred from re-identifying anyone. No method finds everything, and we put no number on our own work until we have measured it.
- 01
Your company controls
What is uploaded. Leave out anything that identifies you by its subject: a product only you make, a public dispute, a deal only you did. Then read the sample and tell us what we missed.
- 02
Mikra controls
The transformation, who sees the raw records, when the raw upload is deleted, and the fix when something is missed.
- 03
Customers are bound
By a contract that bars them from re-identifying anyone.
- 04
If something is missed later
We investigate within 10 business days, correct our copies and future deliveries, and work to have customers replace affected copies.
- 05
Incident notice
If a material security incident affects your data, Mikra tells you within 5 business days.
Data partnerships
Your records,
your review.
Tell us what you hold. We assess it and walk you through this process before you sign anything.