Dan Guido

Most people are now aware that @trailofbits conducted a security review of the Bitcoin Cash client on behalf of @BitcoinSVNode. While we cannot release our report in its entirety yet, I wanted to share a few details of what we found…https://twitter.com/JimmyWinMedia/status/1101668023335833601 …

First, as far as we are aware, this was the first time a professional services firm reviewed the security of a Bitcoin client. We began with a comprehensive review of the bitcoind attack surface and surveyed previous attempts to fuzz it.

In our first week, we developed libFuzzer and AFL drivers targeted to specific areas of high-priority attack surface and modified build scripts to support a “fuzzer mode.” This vastly enhances fuzzing effectiveness, e.g.: https://github.com/google/boringssl/blob/master/FUZZING.md …

These fuzzer drivers and build enhancements were integrated directly into the BitcoinSV codebase, improving its SDLC. Fuzz test corpora are saved after each campaign and milestones are re-fuzzed with these continuously refined test cases. (these are not yet public)

After building fuzzers driven by attack surface modeling and using best practices, we identified a number of security issues over the course of the engagement. This included some of the medium-severity issues that BitcoinSV obtained CVEs to track. https://bitcoinsv.io/2019/03/01/denial-of-service-vulnerabilities-repaired-in-bitcoin-sv-version-0-1-1/ …

These issues detail ways for messages to waste a victim’s CPU and network resources without triggering any of the denial-of-service mitigations that normally detect and ban misbehaving peers.