项目以Big Code背景信息增强作为基本方法、以迁移代码背景语义信息到深度学习反混淆模型为线索，研究海量代码深度背景增强的Android应用代码反混淆，以提升恶意代码检测效率。研究包括基于海量代码迁移分析的混淆代码背景信息增强方法、基于海量代码分析多任务的代码语义模型、海量代码背景信息深度增强的反混淆技术、优化、质量评估、工具和验证。研究贡献在于较早利用涌现的海量代码资源和迅猛发展的深度学习代码分析技术，用于新型代码反混淆研究方法创新和关键技术突破，为发展软件新型技术作出示范；关键技术运用多任务语义建模和背景信息深度增强反混淆，体现了海量代码环境下反混淆研究的独特技术特征。研究意义在于从科技角度尝试回答，能否综合利用现有海量代码资源、代码语义分析和深度学习技术，提供不同于传统的方法和技术手段克服反混淆的内在复杂性。考虑到混淆导致的移动应用高安全风险现状，研究意义明显且具有广阔的推广前景。

Effective code deobfuscation can vigorously promotes malicious code detection for android applications. Therefore our application tends to investigate deobfuscating android applications through Big Code augmented deep learning. The basic method is to take advantage of Android Application Big Code to augmenting the background information of obfuscated codes. It tracks down transfering background semantics of codes to the deep learining enabled deobfuscating model. The investigation includes the method of augmenting the background information of obfuscated codes with Big Code transfering analysis, Big Code multiple tasks of code anlysis driving code semantic model, deobfuscating techniques of Android applications through Big Code Augmented Deep Learning, the optimization schemes, the estimation methods, software tools, and the proof scheme. The contributions of this application will includes that: 1) We early takes advantage of the massive code resources and deep learning enabled code analysis techniques that develop rapidly to innovate new method and breaking through key technologies of deobfuscating codes. This give a good example for innovating software techniques. 2) The implementation of modeling code semantics with Big Code multiple tasks of code anlysis and deobfuscating Android applications through Big Code augmented deep learning, embodies the unique technique characteristics of deobfuscating in Big Code. The scientific interest of the application lies in that we tend to answer the question whether we can comprehensively apply Big Code resources and current techniques such as code semantic analysis, deep learning driven code analysisf to deobfuscate android application codes efficiently in a nontraditional way, by overcoming the interior complexity of deobfuscating codes. Considering the high security risk of android applications caused by code obfuscation, our research has clear academic significance and great practical significance.