Accurate Disassembly of ARMv8-A: Design and Implementation of a High-Fidelity Disassembler

ARMv8-A architecture processors are widely used in high-performance computing, servers, network equipment and scenarios that support large-scale data processing due to their flexibility, low power consumption and high performance. In order to solve the problem of inaccurate existing disassembly tools, this thesis designed and implemented an ARMv8-A disassembler to support the disassembly of three instruction sets: A64, A32 and T32 under the ARMv8-A architecture. The disassembler first collects 9030 valid instructions to implement the function of extracting instruction type decoding information and instruction format information, and then uses the instruction decoder to realize the functions of instruction type decoding and instruction format matching, and finally solves the problem of storing hundreds of millions of undefined instructions through the result exporter. In this thesis, we demonstrate that the constructed ARMv8-A disassembler is more consistent with the results in the ARMv8-A manual than the existing Capstone disassembler by doing accuracy comparison tests.