Dev.to Security 🔐 Cybersecurity 👁 0 📖 6 min read

Why You Can't Hand Your Data to the Cloud

Every few months, a familiar question lands on a tech lead's desk: "Where does this data live? Who can see it? And who changed it?" The first two questions are usually answerable. The third is where most teams get stuck

Every few months, a familiar question lands on a tech lead's desk: "Where does this data live? Who can see it? And who changed it?"

The first two questions are usually answerable. The third is where most teams get stuck.

If your business runs on a cloud SaaS platform or a low-code tool, the honest answer to "where does this data live" is: on the vendor's servers. The customer ledger, the follow-up notes, the configuration you spent weeks tuning — all of it sits in someone else's data center, governed by someone else's policies.

That's fine for a prototype. It becomes a problem the moment a customer contract says "your data must stay in our environment," or a regulator asks for evidence that a record was changed by a specific person at a specific time.

Why it happens

Cloud SaaS is built on a simple trade: you give up control of the data in exchange for not operating it. The vendor hosts the database, patches the stack, and bills you per seat, per month. The price grows with headcount, and the data never comes home.

Low-code platforms make the same trade even more tempting. You can stand up a CRM or an approval workflow in days, with no backend team required. But the speed comes from the vendor's cloud: your business data and your configuration live on their infrastructure, and your only visibility into what happened to a record is whatever the vendor's interface happens to show you.

Three consequences follow:

  • Compliance gaps. If industry regulation or a customer contract requires data to stay in-domain, data on a vendor's cloud is a dealbreaker. You can't sign the contract — or you can't keep the customer.
  • No audit evidence. When an auditor or a customer asks "who changed this record, when, and from where," there is no log you can produce. The vendor may keep logs, but they are not yours to hand over.
  • Lock-in and cost creep. Per-seat pricing means the bill grows with every new hire, and the data that accumulates on the platform is the real moat: leaving means rebuilding the application and migrating years of records.

None of this is a conspiracy. It's just the economics of a managed service. The fix is not to distrust the cloud — it's to make sure the data that can't leave your environment actually stays in your environment.

The foundation: private deployment

Private deployment means the software runs on a server you control, and the database is yours.

With GewuMind Boot, the deployment story is deliberately boring:

  • The distribution package runs out of the box. No Node.js, no npm, no Maven, no Git — you need JDK 17+ and MySQL, and that's it.
  • You run it on your own server — a bare-metal box or a cloud host you manage — and open http://<server-IP>:9089 in a browser.
  • The platform currently works with MySQL 8 only, and all data lives in your own gewumind-boot database.

The platform ships with its own system tables plus menu, role, dictionary, and component data — you create your own database and import them. On top of that foundation, role management covers menus, buttons, and API endpoints with triple authorization, and a data dictionary keeps enum values consistent across the application. Reporting covers list, cross-tab, and Chinese-style reports with conditional formatting, multi-sheet output, merged cells, and detail-with-summary layouts; list pages include built-in Excel export.

The open API gateway is part of the same foundation: each third-party app gets its own appKey/appSecret, with per-app authorization and rate limiting, and every call is recorded.

Licensing is worth understanding, because it shapes the trust model. The source code is Apache License 2.0. Some enterprise capabilities — SSO, workflow, module management, Redis/shared sessions, MQTT, Kafka, SSE, TCP, and WebSocket — are unlocked by a commercial offline license. Without one, you get a 30-day free trial that automatically downgrades to the free base tier when it expires. There are three tiers: Base (free, no license issued), Standard (all commercial capabilities except AI), and Flagship (adds AI customer service, knowledge base, and H5 chat).

Two details matter for a security-minded reader:

  1. License validation is fully offline. A public-key signature check runs at runtime — no network call, no activation server. The license file is bound to the local machine code.
  2. Pricing is per machine, per year — not per seat. Standard starts at 799 CNY / 129 USD; Flagship starts at 1,599 CNY / 249 USD. Your bill does not grow when you hire.

One honest caveat about offline licensing: the trial clock starts on first run on a given machine, and moving the software to another machine or reinstalling restarts the clock. That's an inherent property of offline licensing, not a bug.

Example: three questions, answered

Here is what the audit story looks like in practice, mapped to the three questions from the opening.

"Where does the customer data live?" A sales team runs a customer ledger on the platform: customer records, follow-up notes, and deal stages all sit in your own database, on your server. The records never transit a vendor's cloud, so a contract clause like "customer data must remain in-domain" is something you can actually meet.

"Who changed this record?" The platform keeps two kinds of logs. Operation logs capture key write operations — create, update, delete — with the operator, the time, the IP address, and the affected rows. Login logs record every login attempt: account, IP, time, and success or failure, which is how you spot an abnormal login. When the question "who changed this record" comes up, the answer is a log entry you can produce, not a vendor's word.

"Who is calling our API, and how often?" Third-party systems integrate through the open API gateway. Each app is authorized with its own appKey/appSecret and its own rate limit, and every call leaves an access record: app, endpoint, latency, result code, IP, and time. If an integration starts misbehaving, you can see exactly which app, which endpoint, and when.

Two smaller capabilities round out the operational picture: online session management lets you force a user offline, and built-in database monitoring (Druid) surfaces slow SQL and connection statistics.

What to look for

If you are evaluating private deployment — for this product or any other — here is a short checklist, and two boundaries to keep in mind.

Ask:

  • Where does the data physically live, and whose database is it?
  • Can you produce an audit trail for a specific record change — operator, time, IP?
  • Are third-party integrations governed per app, with rate limits and call logs?
  • How is the license validated? If it phones home, the "private" deployment is only as private as the vendor's network policy.
  • Is pricing per machine or per seat? Per-seat pricing quietly re-creates the lock-in you were trying to escape.

And two honest boundaries:

  • Private deployment is not a managed service. You operate the server yourself, or you need someone who can. The license does not include managed service, custom development, or multi-tenant SaaS rights. If your team cannot run a server, that's a real cost to price in.
  • The product provides data-control capability; it does not replace compliance certification. Classified-protection assessments, industry certifications, and audit opinions remain your company's own responsibility. A private deployment makes those conversations easier — you can point to your own database and your own logs — but it does not certify anything on your behalf.

Try it

If the question "where does this data live" keeps landing on your desk, the fastest way to answer it is to run the software on your own machine and watch the data stay there.

Install the skill, let AI turn requirements into applications — and keep the data where it belongs.

📰 Read the original article on Dev.to Security

Originally published by Dev.to Security. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.