Visible to the public TFCFI:Transparent Forward Fine-grained Control-Flow Integrity Protection

TitleTFCFI:Transparent Forward Fine-grained Control-Flow Integrity Protection
Publication TypeConference Paper
Year of Publication2022
AuthorsShe, Cairui, Chen, Liwei, Shi, Gang
Conference Name2022 IEEE International Conference on Trust, Security and Privacy in Computing and Communications (TrustCom)
Date Publisheddec
Keywordscomposability, compositionality, Computer architecture, control-flow integrity, Hardware, hardware security, Information Reuse, Instruction sets, Law, privacy, pubcrawl, resilience, Resiliency, Rockets, security, Software, system security
AbstractCode-reuse attacks (including ROP/JOP) severely threaten computer security. Control-flow integrity (CFI), which can restrict control flow in legal scope, is recognised as an effective defence mechanism against code-reuse attacks. Hardware-based CFI uses Instruction Set Architecture (ISA) extensions with additional hardware modules to implement CFI and achieve better performance. However, hardware-based fine-grained CFI adds new instructions to the ISA, which can not be executed on old processors and breaks the compatibility of programs. Some coarse-grained CFI designs, such as Intel IBT, maintain the compatibility of programs but can not provide enough security guarantees.To balance the security and compatibility of hardware CFI, we propose Transparent Forward CFI (TFCFI). TFCFI implements hardware-based fine-grained CFI designs without changing the ISA. The software modification of TFCFI utilizes address information and hint instructions in RISC-V as transparent labels to mark the program. The hardware module of TFCFI monitors the control flow during execution. The program modified by TFCFI can be executed on old processors without TFCFI. Benefiting from transparent labels, TFCFI also solves the destination equivalence problem. The experiment on FPGA shows that TFCFI incurs negligible performance overhead (1.82% on average).
DOI10.1109/TrustCom56396.2022.00063
Citation Keyshe_tfcfitransparent_2022