越来越多的现代软件存在测试oracle缺失、大规模的双重特点，同时解决这两个挑战是目前自动缺陷定位领域的空白。本研究旨在提出完整解决方案填补此空白，满足现代软件调试的真实迫切需求。研究内容包括 (1)设计蜕变属性驱动的多层定位框架。首先在数据和算法流程的双重维度进行层次划分，通过逐层选用不同定位算法和数据，满足大规模软件定位高精度低开销的要求，避免传统单一数据分层方法中可能的误差放大。其次将逐层配合定位算法引入蜕变属性概念，使定位技术完全摆脱对oracle的依赖(2)从蜕变属性优选、蜕变切片改进、测试用例集优化三方面研究定位精度优化策略(3)从实用性角度提出综合度量标准，设计高效定位报告提交方式(4)实现工具平台支持上述框架及优化策略，用以实验验证和真实案例分析推广。通过上述研究将极大拓宽自动缺陷定位技术的应用领域，使其不再因程序规模和oracle缺失而受限，从而有效服务于现代软件开发。

Modern software has become more and more large-scaled and most of them have oracle problem. There is a lack of solution that can address these two challenges simultaneously. In this project, we are aiming to propose such a solution in order to meet the real demands in debugging these modern software systems. The research contents include: (1) Design a metamorphic-property-driven multi-level fault localization framework, in which we first build hierarchies in terms of both the spectrum and the localization procedure. Different levels will adopt different data types and analysis algorithms, in order to achieve the corresponding required precision and cost, such that the margin of error will not be amplified when going deeper into lower levels. Then, for each level, we introduce the metamorphic-property in an appropriate manner, in order the make the oracle no longer mandatory in debugging. (2) Design optimization strategies in the perspectives of metamorphic-property selection, metamorphic slice improvement and test suite optimization. (3) Propose a more practical measurement, and design an efficient and helpful report style. (4) Develop a tool-kit to support our framework and optimization strategies, which can be used in empirical analysis and real-life case studies. With the above proposals, the fault localization techniques will become suitable to large-scaled software systems, more importantly, no longer oracle-driven, such that their application domains can be significantly extended and hence can be effectively utilized in real-life development.