Records that hold up. A system we attack every day.
Aeroz keeps custody records, passports and DSCSA cases in an append-only store that refuses edits, signs each audit entry as it is written and seals the log every four hours. We test it daily against real attack techniques before anyone else does.
This page says where Aeroz runs, who can get in, how records stay intact, how we test, who processes your data, and what we don’t claim yet.
One platform, narrow doors.
Where it runs
On Cloudflare. The primary database is in Cloudflare’s Eastern North America region, with no read copies in other regions. Data is encrypted at rest by Cloudflare and in transit with TLS.
Who can get in
People sign in with a one-time code sent by email, with an authenticator-app second factor available. A company that enforces single sign-on can also require the second factor. Each person has a role (owner, admin, manager, operator, viewer or partner) that limits what they see and do.
Keys that can’t outgrow their owner
API keys are stored only as peppered hashes, never readable. Each key carries scopes, cannot mint a key wider than itself, and is revoked when the person who made it leaves.
Limits that can’t be raced
Sign-in attempts, API rate limits and the AI budget are counted with single atomic database writes, so a burst of parallel requests can’t slip past them.
Written once. Checkable forever.
Append-only
Custody records, case histories and the audit log refuse edits and deletes at the database, for Aeroz staff as well as customers.
Signed and sealed
Each audit entry is signed with Aeroz’s records key as it is written, and the log is hash-chained and sealed every four hours, so a back-dated or altered entry is detectable.
Yours to take
Everything exports in open formats (GS1 EPCIS 2.0 and JSON) for an auditor, a regulator or a move to another provider.
We attack it every day.
Each morning an automated run researches attacks seen in the last 48 hours, turns the ones that could touch a system like ours into test scenarios, and replays them against a sandbox copy of Aeroz. A person reviews every fix before it ships.
Built from real incidents
The catalog holds documented incidents from 2023 to 2026, from refrigeration controllers pushed out of range to stolen keys used to run up cloud bills. Each one is replayed step by step against our own code.
A trap for scanners
Requests only an attacker makes, such as hunting for configuration files, are answered from a decoy that never touches real data. A real customer key used from a trapped caller is refused plainly and our team is alerted.
Who processes it, and what never happens to it.
Subprocessors for the platform
Cloudflare, Inc.: hosting, database, file storage and AI models (Workers AI). Resend, Inc.: sign-in codes and notices a person has approved. Nobody else processes customer data in the platform.
On this website
Stripe takes payment for audits and reports. Google Analytics 4 runs only after you accept analytics cookies.
AI, used narrowly
AI triages cases and drafts notices on Cloudflare’s own AI service, so case text stays inside Cloudflare. One model drafts and a model from a different company checks it. A person approves every notice. Your data never trains a model.
No patient data
The platform is built for product and transaction data. Notes that look like they hold patient details are refused, and the same details are blanked out of incoming email and API text.
Said plainly, so you don’t have to ask.
Do you have SOC 2 or ISO 27001?
Not yet. The system is being built toward SOC 2, and there has been no third-party penetration test. We say so here so nobody has to ask twice.
Can Aeroz staff change our records?
No. Custody records, case histories and the audit log are append-only: the database refuses edits and deletes, for Aeroz staff as well as customers. Each audit entry is signed as it is written and the log is sealed every four hours, so a change would be detectable.
Is our data used to train AI models?
No. The models that triage cases and draft notices run on Cloudflare’s own AI service, and the AI log keeps only fingerprints (hashes) of what went in and out, never the text.
How do we report a vulnerability?
Email [email protected]. Our security.txt is published at api.aeroz.io/.well-known/security.txt.
Running a vendor review? Start here.
We answer security questionnaires directly and share our one-pager and subprocessor list with every pilot agreement.