Hardware Mitigation and Verification For Rogue In-Flight Data Load Attacks

Rogue In-Flight Data Load (RIDL) is a microarchitecture security attack that exploits store-to-load forwarding in the line fill buffer. Several microarchitecture-level mitigations have been proposed for defense against RIDL. However, errors and/or Trojans in the implementations of these mitigations can be exploited to render the microarchitecture vulnerable to RIDL. We propose a formal verification methodology that can be used to guarantee that line fill buffer implementations are immune to RIDL attacks. We also propose a hardware-mitigation for RIDL, inspired by our verification approach, that allows the use of store-to-load forwarding. We have demonstrated the efficacy of our approach using several memory pipeline benchmarks.