Browser extensions: Will you ever look at them the same again?

Browser extensions are installed across virtually every workstation in every organisation. Users, including experienced technical professionals, routinely grant them trust based on three signals: the extension comes from an official store, the manifest declares minimal permissions, and the stated purpose appears benign. This presentation demonstrates, with a live working proof of concept, that all three of those trust signals can be systematically undermined.

The session presents a custom-built browser extension whose manifest declares minimal, innocuous permissions. Behind that manifest, the extension contacts a remote command and control server, retrieves a list of target domains to monitor, and when the user visits a monitored site, exfiltrates the full-page content and a screenshot to the operator’s admin interface. The demonstration walks through the permission architecture that makes this possible, the techniques for delayed activation and evasion of static analysis during store review, and the admin interface showing captured data from a controlled test environment.

Recent real-world incidents ground this demonstration in reality. The December 2024 Cyberhaven supply chain compromise affected more than thirty Chrome extensions reaching millions of users. GitLab security research in early 2025 identified sixteen malicious extensions impacting over three million users. Ongoing campaigns have shown extensions operating as sleeper agents, dormant for years before turning hostile. These are not theoretical threats. The permission gap exploited in the demonstration is a live, exploitable weakness in a trust model that practitioners and end users rely on every day.

The session then examines why this works: the structural limitations of browser extension review processes, the gap between static manifest analysis and runtime behaviour, and the asymmetry between what extensions declare and what they can actually do. The defensive discussion covers practical detection approaches, the enterprise controls that meaningfully reduce the attack surface, and the user education challenges that remain stubbornly unsolved.

Crucially, this session does not release exploit code or a functional surveillance tool. The demonstration is conducted in a controlled environment for educational purposes only, and the session adheres to responsible disclosure principles throughout. The goal is to shift how the audience thinks about browser extension trust, not to provide a playbook for abuse.

The presentation is independent research and does not represent any employer or vendor. No product is promoted, and the defensive patterns discussed are applicable to any organisation managing browser extension risk at scale.

Attendees will leave with a concrete understanding of how the browser extension trust model actually fails in practice, and a shifted perspective on the extensions they have already approved in their own environments.