PARALLEL DATA LAB 

PDL Abstract

HerQules: Securing Programs via Hardware-Enforced Message Queues

ASPLOS ’21, April 19–23, 2021, Virtual, USA.

Daming D. Chen, Wen Shih Lim, Mohammad Bakhshalipour, Phillip B. Gibbons, James C. Hoe, Bryan Parno

Carnegie Mellon University

http:/www.pdl.cmu.edu

Many computer programs directly manipulate memory using unsafe pointers, which may introduce memory safety bugs. In response, past work has developed various runtime defenses, including memory safety checks, as well as mitigations like no-execute memory, shadow stacks, and control-flow integrity (CFI), which aim to prevent attackers from obtaining program control. However, softwarebased designs often need to update in-process runtime metadata to maximize accuracy, which is difficult to do precisely, efficiently, and securely. Hardware-based fine-grained instruction monitoring avoids this problem by maintaining metadata in special-purpose hardware, but suffers from high design complexity and requires significant microarchitectural changes. In this paper, we present an alternative solution by adding a fast hardware-based append-only inter-process communication (IPC) primitive, named AppendWrite, which enables a monitored program to transmit a log of execution events to a verifier running in a different process, relying on inter-process memory protections for isolation. We show how AppendWrite can be implemented using an FPGA or in hardware at very low cost. Using this primitive, we design HerQules (HQ), a framework for automatically enforcing integrity-based execution policies through compiler instrumentation. HerQules reduces overhead on the critical path by decoupling program execution from policy checking via concurrency, without affecting security.We perform a case study on control-flowintegrity against multiple benchmark suites, and demonstrate that HQ-CFI achieves a significant improvement in correctness, effectiveness, and performance compared to prior work.

FULL PAPER: pdf