Mitigating Data Leakage by Protecting Memory-resident Sensitive Data

Gaining reliable arbitrary code execution through the exploitation of memory corruption vulnerabilities is becoming increasingly more difficult in the face of modern exploit mitigations. Facing this challenge, adversaries have started shifting their attention to data leakage attacks, which can lead to equally damaging outcomes, such as the disclosure of private keys or other sensitive data. In this work, we present a compiler-level defense against data leakage attacks for user-space applications. Our approach strikes a balance between the manual effort required to protect sensitive application data, and the performance overhead of achieving strong data confidentiality. To that end, we require developers to simply annotate those variables holding sensitive data, after which our framework automatically transforms only the fraction of the entire program code that is related to sensitive data operations. We implemented this approach by extending the LLVM compiler, and used it to protect memory-resident private keys in the MbedTLS server, ssh-agent, and a Libsodium-based file signing program, as well as user passwords for Lighttpd and Memcached. Our results demonstrate the feasibility and practicality of our technique: a modest runtime overhead (e.g., 13% throughput reduction for MbedTLS) that is on par with, or better than, existing state-of-the-art memory safety approaches for selective data protection.