Stage-Supervised Latent Reasoning for Single-Shot JavaScript Deobfuscation

2026-09-22

AI Summary

本文针对JavaScript混淆问题，提出了一种阶段监督的潜在推理框架，通过多阶段重写训练模型以单次生成清晰代码，提高了语法和语义正确性。

Abstract

JavaScript obfuscation is widely used to protect code, but it also makes program analysis and security review substantially harder. Existing LLM-based deobfuscation methods usually treat the task as one-step translation, ignoring the staged structure of practical deobfuscation pipelines. This WIP paper proposes a stage-aware latent reasoning framework that converts intermediate outputs from a deterministic deobfuscation tool into supervision for Coconut-based training. The model learns from multi-stage rewrites during training but generates the final cleaned program in a single shot at inference time. Preliminary results on JsDeObsBench show that the Coconut-based model improves syntactic validity to 50%, compared with 15% for direct fine-tuning and 25% for a zero-shot baseline, and reaches 80% semantic correctness among valid outputs, indicating better semantic faithfulness than either comparison model in deobfuscation.

Problem

• JavaScript obfuscation
• program analysis
• security review
• stage-aware
• latent reasoning

Innovation

• stage-aware latent reasoning
• multi-stage rewrites
• single-shot inference
• Coconut-based model
• syntactic and semantic improvement