This repo contains the benchmarks and analysis scripts used to test Blade in our paper (video here).

We introduce Blade, a new approach to automatically and efficiently eliminate speculative leaks from cryptographic code. Blade is built on the insight that to stop leaks via speculative execution, it suffices to cut the dataflow from expressions that speculatively introduce secrets (sources) to those that leak them through the cache (sinks), rather than prohibit speculation altogether. We formalize this insight in a static type system that (1) types each expression as either transient, i.e., possibly containing speculative secrets or as being stable, and (2) prohibits speculative leaks by requiring that all sink expressions are stable. Blade relies on a new abstract primitive, protect, to halt speculation at fine granularity. We formalize and implement protect using existing architectural mechanisms, and show how Blade's type system can automatically synthesize a minimal number of protects to provably eliminate speculative leaks. We implement Blade in the Cranelift WebAssembly compiler and evaluate our approach by repairing several verified, yet vulnerable WebAssembly implementations of cryptographic primitives. We find that Blade can fix existing programs that leak via speculation automatically, without user intervention, and efficiently even when using fences to implement protect.

The instructions here should allow you to reproduce Table 1 (Section 7) in the paper.

You will need:

• This repo
• lucet-blade from https://github.com/PLSysSec/lucet-blade, on the blade branch. Be sure to check out all submodules as well.
• the WASI SDK from https://github.com/WebAssembly/wasi-sdk
• WABT from https://github.com/WebAssembly/wabt (release 1.0.15 is known to work)
• Binaryen from https://github.com/WebAssembly/binaryen (version 90 is known to work)
• HACL* from https://github.com/project-everest/hacl-star/ (commit de6a314ab is known to work)
• Rust (e.g. from rustup)

In the Makefile in this repo, adjust the variables at the top according to the paths to each of these dependencies on your system.

Also adjust the Cargo.toml in this repo if necessary, with the appropriate path to lucet-blade for the lucet-runtime dependency.

Then, make build in this repo, which should build our modified Lucet, all of our Wasm examples, and the test framework.

In this repo, make bench runs all of our benchmarks. This should take around 10-20 minutes to complete.

(You may see messages about skipped tests, which is normal, or about various outliers, which is also normal.)

Then, use make report to create the table summarizing the results. This outputs a version of the table to stdout, and the actual LaTeX for the table (as presented in our paper) to ./analysis/table.tex.

• After make build, you can inspect the generated x86 assembly for any of our benchmarks by running objdump -SDg on the appropriate .so file in ./wasm_obj.
• You can also inspect the generated WebAssembly for any of the C-language benchmarks by using make wasm_wat/* for the appropriate *.
• To compile other C code with Blade:
  • make build in this repo
  • Compile your code to Wasm using the WASI SDK's clang compiler ($(WASI_SDK)/bin/clang) and the WASI_CLANG_FLAGS and WASI_LINK_FLAGS from the Makefile
  • Then compile these Wasm files to native code using our modified Lucet compiler ($(LUCET_BLADE)/target/debug/lucetc) and the LUCETC_FLAGS from the Makefile.
  • You can choose the Blade mitigation using the --blade-type flag to lucetc:
    • For Ref: --blade-type=none
    • For Baseline-F: --blade-type=baseline_fence
    • For Baseline-S: --blade-type=baseline_slh
    • For Blade-F: --blade-type=lfence
    • For Blade-S: --blade-type=slh
  • You can also choose to enable the v1.1 mitigations with the --blade-v1-1 flag to lucetc (off by default).