Client-Side Zero-Shot LLM Inference for Comprehensive In-Browser URL Analysis

2025-06-04

AI Summary

To address weak generalization, high privacy risks, and vulnerability to advanced adversarial threats in cloud-based malicious URL detection, this paper proposes the first browser-side zero-shot large language model (LLM) security analysis framework. Our method fuses JavaScript abstract syntax trees, dynamic execution traces of DOM/API/network behavior within a sandboxed iframe, and visible page content—enabling real-time, fine-grained maliciousness classification and interpretable threat reporting via multi-source heterogeneous contextual prompting, without parameter fine-tuning. Technically, it integrates WebLLM for lightweight on-device LLM deployment, a genuine browser-based iframe sandbox, and hybrid static-dynamic analysis. Experiments demonstrate that locally executing 3B- or 8B-parameter models achieves detection accuracy comparable to state-of-the-art cloud-based approaches, while providing strong privacy preservation, robustness against anti-analysis techniques, and out-of-the-box generalization across unseen threat variants.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Application Domains: Security
• Natural Language Processing: Safety and Robustness

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• Security and Privacy: Browser and end-point security

Abstract

Malicious websites and phishing URLs pose an ever-increasing cybersecurity risk, with phishing attacks growing by 40% in a single year. Traditional detection approaches rely on machine learning classifiers or rule-based scanners operating in the cloud, but these face significant challenges in generalization, privacy, and evasion by sophisticated threats. In this paper, we propose a novel client-side framework for comprehensive URL analysis that leverages zero-shot inference by a local large language model (LLM) running entirely in-browser. Our system uses a compact LLM (e.g., 3B/8B parameters) via WebLLM to perform reasoning over rich context collected from the target webpage, including static code analysis (JavaScript abstract syntax trees, structure, and code patterns), dynamic sandbox execution results (DOM changes, API calls, and network requests),and visible content. We detail the architecture and methodology of the system, which combines a real browser sandbox (using iframes) resistant to common anti-analysis techniques, with an LLM-based analyzer that assesses potential vulnerabilities and malicious behaviors without any task-specific training (zero-shot). The LLM aggregates evidence from multiple sources (code, execution trace, page content) to classify the URL as benign or malicious and to provide an explanation of the threats or security issues identified. We evaluate our approach on a diverse set of benign and malicious URLs, demonstrating that even a compact client-side model can achieve high detection accuracy and insightful explanations comparable to cloud-based solutions, while operating privately on end-user devices. The results show that client-side LLM inference is a feasible and effective solution to web threat analysis, eliminating the need to send potentially sensitive data to cloud services.

Problem

• Detect malicious URLs using client-side LLM without cloud dependency
• Analyze webpage context via static and dynamic in-browser methods
• Provide zero-shot URL classification with privacy-preserving local inference

Innovation

• Client-side zero-shot LLM inference for URL analysis
• Compact LLM performs multi-source evidence reasoning
• Real browser sandbox resists anti-analysis techniques