Using CSP connect-src to control page network destinations

Core idea

connect-src limits destinations for fetch, XHR, WebSocket, EventSource, and related page connections. A proxy changes the route but not the CSP origin decision, and the browser may block before network use.

Start Using CSP connect-src to control page network destinations with a threat model: what data needs protection, from whom, and on which link. Proxy credentials control access, TLS protects origin traffic, and an IP allowlist narrows the source network. These controls complement one another but do not make an unsafe workflow safe by themselves.

What the primary source establishes

Roll out Report-Only for fetch, WebSocket, and SSE, sanitize reports, and enforce policy without breaking approved integrations.

The primary source, W3C Content Security Policy Level 3, MDN connect-src, MDN CSP Report-Only, WHATWG Fetch, OWASP CSP Cheat Sheet, defines the technical baseline but not every client and provider configuration. Read the normative behavior with its version and then verify your implementation. Treat anything beyond the source as a product feature that needs separate confirmation.

Controlled lab

Call an owned API, WebSocket, SSE, and forbidden endpoint. Collect directive and sanitized blocked origin in Report-Only, test duplicate, oversized, and forged reports, then enforce on staging. Three approved channels must work while the control is blocked before connection.

Step-by-step verification

Start Report-Only on an owned page with a finite list of API, telemetry, WebSocket, and SSE origins. Add one deliberately forbidden destination and review reports before enforcement.

Use a test account and an authorized endpoint. Issue short-lived, least-privilege credentials and define revocation before the run. Verify normal access, rejection with a bad secret, and rejection from an unapproved source. Error messages must remain distinct without echoing credentials.

Inspect application logs, command history, HAR files, and support output. A connection string may leak through a URL, process arguments, terminal history, or screenshots. Rotate temporary credentials and remove sensitive test artifacts after verification.

• Inventory destinations
• Enable Report-Only
• Sanitize and limit reports
• Enforce on staging

Evidence to retain

Keep document origin, effective directive, normalized blocked origin, disposition, browser version, sample count, and owner decision. Remove queries, identifying paths, and cookies.

A security report stores the check ID, time, result, and policy, never a password, cookie, token, secret URL, or reusable browser state. Record that credentials changed rather than storing the old and new values. Evidence should prove the control without becoming another secret store.

Define report columns and time format before the run. A result without context becomes a guess: the address, cache state, and changed condition are unknown. Record controlled failures as well as successes so the check proves that it can distinguish states.

Interpreting the result

Reports can contain URLs and policy data, so limit the collector and remove query details. Do not make a wildcard permanent merely to silence one unexplained request.

CSP reduces browser risk but does not replace server authorization or egress controls. Scheme and reporting support vary by browser.

One successful run confirms only one client, route, and moment. Repeat while changing one variable and state the limits. When observation conflicts with documentation, rule out cache, client version, and intermediaries before creating a reproducible support case.

Worked decision process

Build an access register for Using CSP connect-src to control page network destinations with owner, purpose, expiry, allowed network, and revocation path. Shared credentials prevent selective revocation and useful audit. Issue replacement access, verify one safe request, and then revoke the old secret.

Run negative checks after successful access: old credentials, an unapproved source IP, and an invalid destination. Confirm rejection at the intended stage without echoing a secret. Inspect browser history, process output, CI logs, and support attachments for complete connection strings.

Review active access on a schedule. Match each record to an owner and task, remove orphaned entries, and give exceptions an owner and deadline. The review result is a list of retained, revoked, and time-bounded exceptions rather than a checkbox.

Common mistakes

Disabling certificate validation, sharing one team password, allowing a broad CIDR, and pasting proxy URLs into chat are recurring failures. Restore validation, create scoped credentials, narrow the source range, and verify that old access has actually been revoked.

Stop when errors rise, a source returns a limit, the task would require bypassing protection, or secrets enter logs. Save sanitized diagnostics and correct the cause first. More concurrency or another IP can hide the fault and add load without improving evidence.

Rollout and maintenance criteria

Define the decision boundary before rollout: which observation permits continuation, which requires review, and which stops the workflow. Record acceptable error ratio, maximum wait, and the owner of every exception so a temporary failure cannot silently become permanent configuration.

Review real load, cost, and quality after the first week. Schedule a small control after client, proxy-service, or network changes. Archive outdated instructions with their replacement date and reason so operators do not follow conflicting configurations.

Operational checklist

Turn the successful experiment into a short procedure covering owner, safe configuration, limits, and stop conditions. Every run needs a terminal status. After browser, library, or network changes, run a small control before the main queue.

• Assets and threat are defined
• Test access is isolated
• TLS validation remains on
• Logs were checked for leaks
• Revocation is documented
• Negative cases were tested