Summary
Websites can sometimes behave badly or try to trick your computer, and it's safest to catch these problems right inside your browser before anything happens. The authors built a browser extension called Browser Immune System (BIS) that uses a smart AI model running locally on your computer's graphics processor to quickly check each page you visit for security threats. BIS uses a two-step approach: a fast initial check for known threat types and a slower, detailed explanation only when you want to understand the warning. Their tests show BIS can reliably spot threats on test sites and works similarly across different browsers, though some normal web features still confuse it.
What this means in practice
- •For web browser developers: Integrate AI-driven, on-device security checks in browsers to detect and explain web threats before they affect users.
- •For enterprise cybersecurity teams: Deploy client-side threat detection extensions that provide fast innate alerts and detailed adaptive analyses for suspicious web pages.
Abstract
Security decisions about a web page are best made where the page is: in the browser, before anything leaves the machine. We present the Browser Immune System (BIS), a browser extension that screens navigation with a cybersecurity transformer running entirely on the user's GPU, and alias-web, the model that powers it: a 4B-parameter model that starts from the dual-regime cybersecurity transformer (CT0), extends its public release alias3-micro with a context-category sensor, and is quantised to 4 bits and compiled to a 2.44 GB graph for WebGPU. alias-web is the first in-browser realisation of that model's typed decision mode, System 1, with its official pieces: the marker-token mapping, the complete pointer head and the final normalisation that the exported graph omits. It answers a typed security question in about 150 ms on a laptop GPU, and the same weights generate explanations at 10.5 tokens/s on demand. BIS treats the model as an immune system: an innate response, one System 1 question per threat class, runs on every page; an adaptive response, System 2 text, explains a verdict only when the user asks. On BISBENCH-8, eight synthetic sites of which six execute real but harmless malicious behaviour and two are adversarial controls, BIS returns the expected page verdict on 7/8 sites with both controls green and a class-level micro-F1 of 0.71, scans a page in 0.9 to 6.4 s, and gives identical verdicts across samples and across Chrome and Firefox. Single-variable diagnostics trace both benchmark errors to what the model was shown and, with a first pass over eleven real public pages, show that two of the seven questions answer yes on ordinary scripts and login forms: specificity on the real web is the open problem. We report every error, the engineering obstacles between a research checkpoint and a working extension, and what the evidence does not yet support.