Eleuther Browser

⚠️ Note: This project is not intended as a full-featured daily driver browser.
It is a security-focused experimental wrapper around QtWebEngine.

GitHub: https://github.com/Dev-Devon/Eleuther

---

Overview

Eleuther is a prototype browser security monitor built on top of QtWebEngine.

Instead of trusting all web content by default, it introduces a policy-based interception layer that evaluates network requests before they are executed.

The goal is to explore stricter control over web execution behavior, especially around third-party scripts and telemetry endpoints.

---

Core Design

1. Request Interception Layer

Every outgoing request is processed through a PolicyEngine that returns a structured decision:

• ALLOW
• BLOCK

Each decision includes:

• Action
• Confidence score
• Reason for decision

This allows inspection and debugging of why specific requests are blocked.

---

2. Domain Filtering

The system maintains two primary domain categories:

• Blocked domains: commonly used ad and tracking networks
• Telemetry-scoped domains: domains where additional path-level filtering is applied

Requests are evaluated using both hostname and path-level matching.

---

3. Path-Level Filtering

Certain request patterns are blocked when detected on sensitive domains:

• tracking endpoints
• analytics beacons
• pixel-based telemetry routes

This is applied only to specific resource types (e.g., XHR, Script, Fetch).

---

4. DOM Sanitization Layer

A lightweight JavaScript injection runs after page load to remove known advertisement-related elements.

It uses:

• attribute-based selectors ([is-promoted])
• class-based selectors (YouTube UI ad containers)
• iframe filtering for known ad sources

It is triggered on mutation events rather than polling intervals to reduce CPU overhead.

---

Current Capabilities

• Request interception with policy-based evaluation
• Domain and path-level blocking
• Lightweight DOM cleanup injection
• Basic tabbed browsing via QtWebEngine
• Persistent profile storage (non-incognito mode)

---

Limitations

• Performance overhead due to Python + QtWebEngine integration layer
• Aggressive filtering may break authentication flows or third-party widgets
• DOM sanitization depends on stability of external platform markup (e.g., YouTube structure)
• No adaptive learning or heuristic model currently implemented

---

Potential Improvements

Native Performance Layer

Porting the network filtering logic to a native extension (Rust or C++) would reduce latency and improve scalability under heavy request loads.

Heuristic Filtering

Current filtering is rule-based. A future version could incorporate:

• anomaly detection for request patterns
• adaptive classification of telemetry endpoints

DOM Resilience

The DOM cleaning system is brittle by design and depends on static selectors. A more robust approach would require:

• structure-aware heuristics
• mutation diff classification
• or integration with browser-level APIs

---

Contribution Areas

Help is needed in the following areas:

• Rust/C++ developers for native network filtering module
• Web security researchers for improving heuristic accuracy
• Frontend engineers for improving DOM sanitization resilience
• Performance optimization of QtWebEngine integration layer

---

Disclaimer

This project is experimental.

It is not a replacement for a production browser and may break websites due to its restrictive filtering behavior.

---