Injected plaintext redirects
Google TAG documented an Intellexa campaign in which an intercepted HTTP visit silently redirected a selected target to an exploit server. Blocking top-level HTTP removes that exact transport opportunity.
Targeted-attack browser hardening · optional · off by default
Two bounded request controls for higher-risk browsing.
ZSEC Browser Shields Community 0.4 can block two documented browser-delivery paths before matching requests are sent. It does not inspect an iPhone, detect Pegasus or prevent an unknown vulnerability.
Attribution boundary: Pegasus is publicly attributed to NSO Group, an Israeli company. That is different from claiming every commercial-spyware campaign is Pegasus or attributing a specific attack without forensic evidence.
What independent research shows
Browser-delivered attacks are only part of the threat model.
Commercial surveillance operators have used drive-by links, injected redirects, compromised websites, hidden frames and multi-stage browser exploits. Other attacks need no browser visit or click at all.
Watering-hole frames
Google TAG documented compromised government sites loading hidden third-party iframes and exploit code closely related to tooling previously used by commercial surveillance vendors including NSO Group.
Zero-click mobile chains
Citizen Lab's BLASTPASS investigation described an NSO Group exploit delivered through PassKit attachments sent via iMessage. Earlier chains targeted HomeKit, Find My and iMessage. A desktop Chromium extension is not in those paths.
Updates remain fundamental
Researchers repeatedly found exploit chains against devices that had not installed available security fixes. Extension rules add a layer; they do not patch Chromium, WebKit, a kernel, baseband firmware or another application.
Enforced in Community 0.4
Two local rules. Two clear decision points.
Both the normal protection switch and High-Risk Browsing must be on. The browser's Manifest V3 engine evaluates the fixed rules locally before a matching network request is sent.
Rule 200000 · plaintext main frames
Blocks top-level URLs beginning with http://. It does not silently upgrade or bypass the decision. HTTPS prevents plaintext network injection but does not certify the destination as benign.
Rule 200001 · third-party active delivery
Blocks third-party script, sub_frame, object and websocket requests. First-party content and unlisted resource types remain outside this rule.
Runtime evidence—not only unit tests
On 21 August 2026, disposable tests in Brave 1.93.137 and Edge 151.0.4129.101 on Windows loaded the unpacked extension into fresh temporary profiles. The tests proved the exact rules were installed, a third-party script and top-level HTTP navigation were blocked before reaching loopback servers, and explicitly disabling High-Risk Browsing removed both rules and restored navigation.
The Chrome 152 command-line harness could not load an unpacked extension, so Chrome is not included in that runtime-smoke claim. Manual Manifest V3 compatibility is a separate install path.
The rules have priority 1000, above the normal site-pause allow priority. The popup disables site pause while the profile is active. A HIGH badge means the two local settings were successfully applied; it is not a threat or infection verdict.
Non-negotiable limitations
What remains possible.
- An exploit embedded directly in first-party content may still run.
- Images, media, stylesheets, fonts, fetch/XHR and other permitted channels may still carry hostile data.
- An unknown browser, extension, kernel, driver, operating-system or hardware vulnerability may bypass this layer.
- A zero-click message, call, push service, HomeKit, Find My or baseband route may never enter Chromium.
- A compromised extension or local administrator can change the wider security outcome.
- No block means only “no configured rule matched”—never “safe” or “clean.”
ZSEC decision pointBlock matching HTTP navigation and third-party active requests
Browser-managed protectionsUpdates, Safe Browsing, sandbox and Site Isolation remain authoritative
Operating-system protectionsNative security, primary antivirus and platform mitigations remain active
Outside this extensionMessages, memory, kernel, baseband, first-party and permitted content
No layer is presented as absolute protection
High-risk operating guide
Keep browser, account and platform security controls enabled.
High-Risk Browsing is one optional desktop-browser layer. The device vendor remains responsible for patching and its privileged security controls.
Every desktop
Install browser and operating-system updates promptly, restart to complete them, keep Safe Browsing or the vendor equivalent enabled, and never bypass certificate or download warnings.
Microsoft Edge
Microsoft's Enhanced Security mode can disable JIT on unfamiliar or all sites and enable additional operating-system mitigations. Strict mode may break normal work, so treat it as a separate high-risk choice.
Apple targeting alerts
Verify an Apple threat notification by signing in directly at account.apple.com. Apple recommends Lockdown Mode and qualified expert help for notified or credibly targeted users.
Credible targeting concern
Stop interacting with the suspect content, preserve the device state, use a separate trusted device for help, and contact a qualified incident-response channel. Do not “clean up” evidence or treat an antivirus scan as forensic clearance.
Access Now Digital Security Helpline · Amnesty MVT limitations
Inspectable protection
Verify the bytes, then choose the trade-off.
Community 0.4 is unsigned and manually updated. Check the published SHA-256, inspect the immutable source and test it in a separate browser profile before relying on it.