Spot disease outbreaks
before they spread.
Privacy-preserving disease surveillance for African public health agencies. Built on anonymised signals from the Therasafety consumer platform. k-anonymity ≥5, differential privacy. Never identifies individuals.
- NCDC-aligned
- k-anonymity ≥5
- Differential privacy
Nigeria detects outbreaks
14 days late.
By the time NCDC sees a confirmed lab report, the disease has been spreading for two weeks. Cholera. Lassa. Mpox. Yellow fever. A 14-day blind spot is the difference between a contained outbreak and a national emergency.
Median detection lag
From first symptom presentation in a community to NCDC’s national dashboard. By that point, secondary cases are already infected and uncounted.
Manual paper reporting
Most state surveillance still depends on weekly paper case forms travelling through hierarchical channels — when they make it through at all.
Real-time signal layer
No symptom-level visibility into communities where most disease starts. Surveillance only sees confirmed lab cases — far downstream from where action would matter.
From millions of consumer signals
to early outbreak detection.
Consumer products contribute signals.
When a HowBody user mentions GI symptoms in Pidgin, when a WetinDey user uploads a malaria-positive lab result, when a BodiCheck device records elevated fever — each becomes an anonymous signal.
⌘ 8 consumer products feed Vigil
Privacy preservation applied first.
Before any signal reaches Vigil, k-anonymity (≥5 individuals per query group) and differential privacy noise are applied. Identifiers are stripped. Locations are coarsened to LGA level.
⌘ Privacy-preserving by construction
ML detects unusual patterns.
Vigil’s anomaly detection compares current signal density to baseline patterns. Unusual clusters trigger alerts to NCDC and state ministries — in hours, not weeks.
⌘ Hours-to-alert, not days
We never identify individuals.
The single most important design constraint of Vigil is this: useful surveillance signals can be extracted without ever knowing who any given person is. This is technically achievable — and Vigil is built on that principle from the foundation up.
Identifier stripping at the edge
Before any signal leaves a consumer product, all direct and quasi-identifiers are removed. NaijaHealth ID, phone number, name, exact location, exact timestamp — all stripped at source. Only the symptom or condition signal travels onward.
// PII never reaches Vigil’s databasek-anonymity ≥ 5 enforced at query
No query result can return a group smaller than 5 individuals. If a query for "fever cases in Ward 12, last 24 hours" would return only 2 records, Vigil widens the query (broader ward, longer window) until k≥5 is satisfied — or refuses to answer.
// SELECT only when count(*) ≥ 5Differential privacy noise injection
Calibrated random noise is added to all aggregate counts. An attacker comparing yesterday’s count to today’s count cannot infer whether any specific individual was added or removed. Privacy budget (ε) is bounded and audited.
// ε ≤ 1.0 per epoch · Laplace noiseGeographic coarsening
Locations are reported at LGA level (774 nationwide), never at street, ward, or community level for low-density alerts. The trade-off is small for surveillance accuracy and large for privacy protection.
// Min granularity: LGA · 250k+ pop.Vigil’s privacy architecture is open to external audit by privacy researchers, NCDC, and the National Data Protection Commission. We publish our privacy budget, query patterns, and aggregation methodology.
Built for public health credibility.
Vigil is the most sensitive product in the Therasafety platform. Every architectural decision reflects that. We work alongside the regulators, not around them.
NCDC partnership-first
Vigil isn’t a parallel surveillance system. It augments NCDC’s existing infrastructure with real-time community-level signals. All anomaly alerts route through NCDC’s epidemiology team for clinical interpretation before public action.
Privacy is the architecture
Privacy isn’t a policy bolted onto Vigil — it’s how the system is built. k-anonymity, differential privacy, identifier stripping, geographic coarsening. Engineering invariants enforced in the codebase, not in handbooks.
Open methodology, audited code
The privacy mechanisms underpinning Vigil — query rewriting, noise calibration, k-anonymity enforcement — are documented publicly and available for academic and regulatory audit. No security through obscurity.
Request a technical briefing.
Vigil is in pre-launch development with NCDC and African public health agencies. If you represent a national or state-level health authority, request a technical briefing on the privacy architecture and integration timeline.
- For NCDC, Africa CDC, and SMOH partners
- Privacy whitepaper available on request
- Pilot integration starting Q3 2029
Vigil is built on 8 consumer products that millions of Nigerians use.
Partner with us to protect
African public health.
Vigil is in pre-launch development with NCDC and African public health agencies. Request a technical briefing on the privacy architecture and integration timeline.