Séminaire confiance numérique (Catalin Hritcu, Equipe INRIA PROSECCO)

Efficient Formally Secure Compilers to a Tagged Architecture.

Severe low-level vulnerabilities abound in today's computer systems, allowing cyber-attackers to remotely gain full control. This happens in big part because our programming languages, compilers, and architectures were designed in an era of scarce hardware resources and too often trade off security for efficiency. The semantics of mainstream low-level languages like C is inherently insecure, and even for safer languages, establishing security with respect to a high-level semantics does not guarantee the absence of low-level attacks. Secure compilation using the coarse-grained protection mechanisms provided by mainstream hardware architectures would be too inefficient for most practical scenarios.


This talk will present a new 5 year project aimed at leveraging emerging hardware capabilities for fine-grained protection to build the first, efficient secure compilers for realistic low-level programming languages (the C language, and Low* a safe subset of C embedded in F* for verification). These compilers will provide a secure semantics for all programs and will ensure that high-level abstractions cannot be violated even when interacting with untrusted low-level code. To achieve this level of security without sacrificing efficiency, our secure compilers target a tagged architecture, which associates a metadata tag to each word and efficiently propagates and checks tags according to software-defined rules. We hope to experimentally evaluate and carefully optimize the efficiency of our secure compilers on realistic workloads and standard benchmark suites. We are also using use property-based testing and formal verification to provide high confidence that our compilers are indeed secure. Formally, we are constructing machine-checked proofs of fully abstract compilation and of a new property we call robust compilation, which implies the preservation of safety properties even against an adversarial context. These strong properties complement compiler correctness and ensure that no machine-code attacker can do more harm to securely compiled components than a component already could with respect to a secure source-level semantics

501 vues
Jeudi 06 Avril 2017
Tag(s) : confiance numerique, sécurité informatique
Partager : Facebook
Intégrer :
Vidéos suggérées