Architecture and Data Security
Last updated November 16, 2026
1. Why this document exists
ISL Data was built and is managed by a core development and operations team. This statement covers:
- what the system is made of,
- where data physically lives,
- the threats we consider most, and what we do about each one,
- the role of our system development and maintenance team,
- what it means if you host the system yourself.
Separate from these protocols and practices, our Terms of Service govern our assumption of risk, liability, and your rights to your data.
2. Who we are
ISL Data is built bit by bit and run by a small, dedicated team of researchers and engineers. As a user you have direct access to the people who build the system. Good data security is a shared practice: we see it as a data culture we build together with the projects we serve.
3. What ISL Data is, and where data are held
ISL Data is a data capture, cleaning, management, analysis, and reporting system. The moving parts:
| Component | What it does | Where it runs (ISL-hosted) |
|---|---|---|
| Data Viewer application | Web app for access control, import, review, edit, create, analyze, and audit functions | DigitalOcean (region: SFO3) |
| MongoDB database | Stores records, users, views, audit log, sessions | DigitalOcean (region: SFO3) |
| Object storage | Photos, audio, and other binary attachments | Amazon S3 (region: us-west-1) |
| Data system integration | A variety of third-party digital collection tools that feed into the database (ODK Central, ESRI Survey123, Google Forms, KoBoToolbox) | Each platform runs on its own host, operated by its vendor or by you — not by ISL (see below). |
Resources we integrate with. ISL Data currently facilitates automatic data retrieval from Open Data Kit (ODK), Google Forms, ESRI Survey123, and KoBoToolbox, and accepts standard spreadsheet and CSV files. The integrations retrieve data into a designated ISL Data workspace; they do not establish a live link to remove, edit, or archive anything in your data collection platform. The use of MongoDB for ISL Data allows the system to be highly flexible and source-agnostic.
The security of data before it reaches ISL Data is governed by the data collection service vendor — Google for Google Forms, ESRI for Survey123, your KoBo server's operator. We do not enforce access control, retention, or deletion inside those systems. If your data is sensitive, consider collecting with encrypted forms, which most data collection apps allow.
4. Data security
Authentication
ISL Data uses passwordless email sign-in. When you navigate to the database, the system requests an email address. If the email address matches a registered user, a single-use, signed link is sent to that email; links expire after 5 minutes. Following the link establishes an ISL Data browser session. Passwordless email sign-in means there is no password theft and no account sharing, unless via a shared email account. It ensures that users can quickly enter the system, no matter how infrequently they need it, with an email account they use every day.
For standard users, an email inbox is the credential. Anyone with persistent access to a registered user's mailbox can sign in to ISL Data as that user; however, damage is limited to that user's access: they can see and alter only objects they have been granted access to, and all their actions are recorded in the audit log that administrators can see. We highly recommend that every ISL Data user enable multi-factor authentication (MFA) on the email account associated with ISL Data.
If you require an additional password entry mechanism, please contact support at info@insitulabs.org.
A browser session stays active while you use it and ends after 5 days of inactivity; then you must sign in again. Every sign-in attempt — including attempts using email addresses that aren't registered — is recorded in the audit log with the IP address and browser used.
Authorization — workspaces and per-object access control
Data is isolated into workspaces, each presenting a separate database. A user is a member of specific workspaces and sees nothing outside them. Within a workspace, access is granted by administrators to each individual object — per source, per view, per report, per dashboard — with separate read and write permissions. A user can be given write access to exactly one data source and nothing else.
Access is default-deny: new users see nothing until granted access by a workspace administrator.
Encryption
- Data in transit: connections to the application and to S3 storage are over HTTPS/TLS. Collection apps and platforms connect over HTTPS.
- At rest: database storage and object storage are encrypted at the provider level — AWS S3 server-side encryption and DigitalOcean volume/managed-database encryption.
Attachments
Photos and other media are stored in a private AWS S3 bucket and are never served from a public URL. Each request generates a short-lived pre-signed link (default 60 seconds) scoped to that one object. Copying an image URL out of the app does not create a shareable link.
Audit logging
The built-in audit system records:
- object visits;
- logins and login attempts;
- data exports and downloads;
- record creation, edit, delete, and restore;
- imports and import edits;
- creation, edit, deletion, and restoration of sources, views, and reports;
- user permission changes.
The log keeps the previous value of every changed field, so any record edit or import can be undone from the audit log by an administrator. Changes to attachments (photos, audio) are the exception: edits or deletions cannot be undone in this way. Audit records are append-only through the web application; they are not editable through the user interface.
Operational practices (ISL-hosted)
- Named individual accounts on DigitalOcean and AWS, MFA enforced, no shared logins, no root/console password sharing.
- Application credentials (database URI, S3 keys, token signing secrets, email-service credentials) held as environment variables outside the code repository; rotated on staff change and on suspected exposure.
- S3 access via a least-privilege IAM user scoped to the single ISL Data bucket, not an account-wide key.
- Firewall restricting the database to the application host only; no database port exposed to the internet; administrative access via SSH key only, password authentication disabled.
- Operating system and dependency patching on a defined cadence; dependency vulnerability scanning in the build.
- Automated encrypted backups: 7-day point-in-time recovery and 90-day daily snapshots (see Section 5), with restores tested within the last 12 months.
- Application error and access logging, reviewed on alert.
5. Data retention, deletion, and portability
- Your data is accessible to you at all times. Export it in full at any time in open formats (CSV, spreadsheet, attachments).
- Deletion. The system does not delete; it archives. Archived records do not appear in views, dashboards, or reports, but are restorable indefinitely. If workspace size becomes prohibitive due to archived records, contact support for hard-deletion support.
- On termination, the system renders the full workspace in a compressed file format that remains downloadable by administrators for 30 days. The export is deleted after 30 days, or sooner upon request. Copies in routine backups expire within 90 days.
- Backups. We keep two layers of automated backups. Continuous backup (7 days): changes are captured continuously, so within the last 7 days we can restore to any specific point in time, limiting data loss after an error or incident to the moments just before it. Daily snapshots (90 days): a full snapshot is taken every day and kept for 90 days, so we can recover data from a known-good state up to three months back.
- Restores. A single workspace, including its attachments, can be restored without affecting other workspaces. Requests within the past 7 days can target any point in time; requests from 8 to 90 days ago are restored from the nearest daily snapshot. Data older than 90 days cannot be recovered from backups. To request a restore, email info@insitulabs.org. Backups are a safeguard, not a substitute for your own exports.
6. Solutions to common threats
| Threat | Realistic scenario | System solutions | Your responsibility |
|---|---|---|---|
| Stolen or borrowed credentials | A collaborator's email account is compromised, or a shared field email is used by several people | Passwordless links expire in 5 min and are single-use; sessions expire after 5 days of inactivity; all logins audited | Multi-factor authentication (MFA) on every user's email; one account per person; do not leave stale users on the workspace |
| Lost or stolen field device | A tablet or phone with unsent form submissions is lost | Digital form data already submitted is secured by the third-party collection service and additionally by ISL Data | Device passcode + full-disk encryption; submit digital forms as soon as possible |
| Over-broad internal access | Everyone in an organization can see and edit the data of every internal group | Complete workspace isolation; per-source/view/report read and write permissions; default-deny access for new users | Grant minimum permissions; review the user list regularly; keep the number of workspace admins small |
| Accidental or malicious data damage | A bulk edit or import overwrites good records | Full audit trail; imports and edits are reversible; archive-only with full restore capability; automatic backups | Use views rather than editing sources where possible; inform the ISL support team quickly |
| Data leaving via export | A CSV of everything ends up on a personal laptop or in a shared drive | Exports and downloads are logged with user and time; attachment links expire | Set an internal rule for where exports are saved |
| Server compromise via unpatched software | A known vulnerability in a dependency or the OS is exploited | Patch cadence, dependency scanning, minimal exposed surface, no framework version disclosure, firewalled database | Nothing, if we host. Everything, if you host — see Section 9 |
| Temporary loss of service | The application is flooded and becomes unavailable | Provider-level network protection; this is an availability risk, not a confidentiality one | Use of third-party collection tools means collection is unaffected; sync resumes when service is restored |
| Third-party source systems | Data is exposed inside a collection platform we do not operate, before it ever reaches us | Outside our control; however, ISL Data retains a secure copy within the controls outlined in this statement | Secure those accounts under that vendor's terms; for sensitive projects, encrypt submissions if possible |
| Insider access by ISL staff | An ISL super administrator reads your data | See Section 8 | Pick a hosting model that matches your requirements |
7. Recommendations for all users
Security is a shared responsibility. Each recommendation below is in your control and will reduce risk:
- Require MFA (two-factor authentication) on all email accounts associated with your ISL Data workspace.
- One account per person. Do not allow shared user email accounts.
- Least privilege. Grant per-source and per-view access, not blanket administrator rights. Keep admin accounts to the minimum.
- Off-board promptly. Remove the user the same day someone leaves the project.
- Device hygiene for field devices: require passcodes, encryption, an updated OS, and an admin/settings password on collection apps when supported.
- Decide where exports may live. Exports are the boundary of our auditing system.
- Review your user list and audit log regularly.
8. Access by ISL system administrators
Two levels of "administrator"
- Workspace administrator — an individual who has full rights to view, create, manage, and audit within a specific workspace. ISL Data subscribers choose who these individuals are.
- ISL system administrator ("super-admin") — ISL technical staff who operate the platform: the servers, the database, the object storage, the deployments, the backups.
When subscribed to an ISL Data hosted plan, technical staff are able to access your workspace to provide critical technical support. A system that is completely inaccessible to our team requires self-hosting; refer to Section 9 to learn about the security and risk consequences of self-hosting. Our Terms of Service (Section 4.4) define that system administrators are prohibited from reviewing, altering, exporting, or sharing any aspect of your data without data-security cause or unless expressly requested by you. In rare cases, such as a subscriber introducing malware that would compromise their workspace as well as others, system administrators will inform subscribers and take immediate safety actions in accordance with the severity of the data threat. For any other reason — e.g. poor system performance related to your workspace — system administrators must notify you and request permission to inspect and correct any issues. When granted permission, an ISL system administrator can:
- open a workspace in the application with full administrative rights;
- query the database directly, which reaches all records, all users, and the audit log;
- read attachments in object storage;
- read and change configuration, deploy code, and access backups.
These capabilities are necessary to ensure the system works reliably for you and for all other users. Except in extenuating data-security circumstances, your data are accessed only when approved, by core technical staff, at times that do not interfere with your operation, and with an issue completion report provided to you.
Functions of the system administrators
- Provide documented support that is requested — you report a problem that cannot be diagnosed without looking at the records.
- System-wide operations and functionality — migrations, backup restoration, sync repair, performance investigation, integrity checks, functional improvement. These are structural (schema, counts, indexing) maintenance and improvement tasks, and do not expose data.
- Security incident — investigating suspected compromise or misuse.
- Legal obligation — a valid legal order. We will notify you unless legally barred from doing so.
System administrators may not:
- browse project data out of curiosity;
- use subscriber data in any analysis, publication, or reporting;
- share or sell subscriber data;
- use subscriber data to train machine-learning or AI models;
- grant access to funders, partners, or other ISL Data users outside your workspace without your written instruction.
Is administrator access recorded?
All in-app actions by global administrators are recorded. Once you give approval, you will see an ISL system administrator added to your user list during an active support request, and the system will record their actions. Actions taken by an ISL administrator who is a member of your workspace are audited exactly like any other user's. This is the standard practice, with one exception: "Guest" mode.
A "Guest" mode exists for system administrators for detailed repair or system debugging without polluting project audit history. The effect is that you will not see system administrator actions while in Guest mode. Our staff will always contact you before accessing your workspace, except in extenuating circumstances (e.g. major security threats). Supplemental audit logs of system administrator activity associated with your workspace are available on reasonable request. In these instances, database and object-storage access is outside application logging but does appear in infrastructure/provider logs, which can be provided on reasonable request.
9. Self-hosting
Self-hosting may be necessary when data-sovereignty requirements, sensitive human-subjects data, or government or institutional policy does not allow external support or access under any circumstance. Self-hosting is also common for a university, agency, or other institution with its own advanced IT team. In this case, code to run the ISL Data platform is available for a one-time, single-use licensing fee.
If you are considering this option, please be aware of the following technical responsibilities — not only at installation, but also as IT environments evolve over time.
- Patching. OS, Node.js, MongoDB, and application dependencies. Unpatched systems are the most common cause of compromise. This is a recurring duty.
- Upgrades. Tracking ISL Data releases, including those that fix security issues, and applying them.
- TLS certificates. Issuance and automatic renewal. Expiry causes visible outages and trains users to click through warnings.
- Backups and tested restores. Database and object storage, encrypted, off-host, with a restore you have actually performed.
- Access control on infrastructure. SSH keys, firewall rules, MFA on your cloud accounts, off-boarding.
- Secrets management. Token signing secrets, database credentials, S3 or compatible storage keys, email-service credentials — stored outside the repository and rotated.
- Email deliverability. Sign-in depends on email. Misconfigured SPF/DKIM/DMARC means links land in spam and nobody can log in.
- Object storage configuration. A private bucket, correctly configured. A public bucket is the most common and most damaging misconfiguration in this class of system.
- Monitoring. Noticing that the system is down, that sync stopped, that the disk is full, or that something unusual is happening in the logs.
- Incident response. If it is compromised, you detect, contain, investigate, and notify.
An unmaintained self-hosted instance is less secure than a maintained hosted one, and it degrades silently. If no named person at your institution owns this work with budgeted time, self-hosting is a high-risk option. Institutional IT departments with reliable capacity are an excellent reason to self-host.
Middle options (for enterprise users). On a case-by-case basis, we will consider: (a) hosting in a region or provider account you specify; (b) deploying into infrastructure you own while we operate it under agreement; or (c) splitting the hosting — for example, you self-host ISL Data while we run your data collection server, or the reverse. If neither standard option fits your institution's policy, ask.
10. Incident response and vulnerability reporting
If we have a security incident affecting your data, we will tell you what happened, what data was involved, what we have done, and what you should do — without undue delay and in any case within 72 hours of confirming it — while we remedy the situation as quickly as possible.
To report a vulnerability, email data@insitulabs.org.
11. Where we're heading
Where we are today
- Certifications such as SOC 2 and ISO 27001 are designed for much larger companies, and we will pursue them as our user base grows and requires it.
- ISL Data is compatible with, but not currently credentialed for, protected health information (HIPAA). Tell us early if your project involves human-subjects data.
- We are committed to continuous uptime, and we welcome your questions and concerns so that we may address them directly.
What we are improving next
- System penetration testing.
- Application-level limits on repeated sign-in requests (Section 4).
- Structured approval processes for ISL support requests (Section 8).
- Workspace administrator control of additional password protection.
12. Our commitment
We review our security practices at least annually, tell you when something material changes, and welcome your questions and suggestions.