← All expertise

04

Security and observability

A secure system is not one without incidents: it is one where the incident is visible, explainable and evidenced. All three have to be built before you need them.

Noticing before the customer does

Metrics exposed by the services, health checks that tell "powered on" from "working", and alerts that fire on useful symptoms instead of every flicker. An alarm that always rings is an alarm switched off.

Where it makes sense the reaction is automatic and graded: try the light remedy, then the heavy one, and if the problem repeats, open the circuit instead of restarting forever — a breaker that stops insisting is more useful than one that reboots endlessly.

Strong access and secrets that never sit in the repo

Mandatory second factor for administrators, passkeys where users can actually use them, short-lived internal certificates, credentials generated and renewed by a vault. Keys pasted by hand into a chat are burned and must be rotated: a boring rule, and a necessary one.

Backups you can actually restore

An unverified backup is just a file. The restore drill must be automated onto a throwaway database and must also check that the protections really come back — not that they exist, that they work: if the role returned with the wrong privileges the policy count would look identical while every customer silently saw everyone else's data.

Portfolio

What I did on this, project by project.

Orbitas

Managed security of the private cloud

  • A Wazuh agent on every virtual machine with per-customer isolation: vulnerability detection, file integrity monitoring, event correlation.
  • Automated incident response across three escalation levels with a circuit breaker and error classification, keeping a full trail of what was attempted and when.
  • TLS everywhere from Vault's PKI, second factor for operators, management network over a WireGuard overlay.
  • Append-only event log as the basis for compliance evidence, plus security score and incident history exposed to the customer.
  • Wazuh SIEM
  • Prometheus
  • Grafana
  • Consul health
  • Vault PKI
  • WireGuard
  • audit log

SOS Bollette

Continuity of the multitenant CRM

  • Automated weekly restore drill: it rebuilds the latest backup onto a temporary database, verifies its SHA-256 fingerprint before touching anything, then drops it.
  • The drill does not count policies: it connects with the application role without declaring an actor and checks that no row comes back — the only test that exposes a role restored with the wrong privileges.
  • No command can overwrite the live database: a restore creates a new one, so if it doesn't convince you there is still a way back.
  • PostgreSQL
  • pg_dump / pg_restore
  • RLS
  • cron
  • SHA-256

Manora

Security of a platform that moves value

  • Static analysis over every contract with Slither and Mythril in the pipeline, with documented results for external auditors.
  • Threat model covering actors, trust boundaries and fund flows; an emergency runbook with response procedures written before the incident.
  • A dedicated exporter bringing chain state into Prometheus and Grafana alongside application metrics.
  • Slither
  • Mythril
  • threat model
  • runbook
  • Prometheus
  • Grafana

Need this?

Tell me the problem and I'll tell you how I would tackle it — and if it isn't worth doing, I'll tell you that too.

Let's talk