Witnessing Secure Compilation

Transformations everywhere

Analyzing and optimizing program memory performance is a pressing problem in high-performance computer architectures. Currently, software solutions addressing the processor-memory performance gap include compiler-or programmer-applied optimizations like data structure padding, matrix blocking, and other program transformations. Compiler optimization can be effective, but the lack of precise analysis and optimization frameworks makes it impossible to confidently make optimal, rather than heuristic-based, program transformations. Imprecision is most problematic in situations where hard-to-predict cache conflicts foil heuristic approaches. Furthermore, the lack of a general framework for compiler memory performance analysis makes it impossible to understand the combined effects of several program transformations.The Cache Miss Equation (CME) framework discussed in this paper addresses these issues. We express memory reference and cache conflict behavior in terms of sets of equations. The mathematical precision of CMEs allows us to find true optimal solutions for transformations like blocking or padding. The generality of CMEs also allows us to reason about interactions between transformations applied in concert. Unlike our prior work, this framework applies to caches of arbitrary associativity. This paper also demonstrates the utility of CMEs by presenting precise algorithms for intra-variable padding, inter-variable padding, and selecting tile sizes. Our experiences with CMEs implemented in the SUIF system show that they are a unifying mathematical framework offering the generality and precision imperative for compiler optimizations on current high-performance architectures.

A lot of programming models have been proposed to deal with parallelism in order to express program transformations and refinements. This justifies to introduce an unifying theory to abstract different concepts. The aim of this paper is to introduce such a theory. This theory includes the definitions of problems, programs and transformation rules. It is founded on the simple mathematical concepts of multiset and of an equivalence between their representations as data fields. Program transformations are founded on this equivalence and defined from a refinement relation. Due to the unifying aspect of this theory, solutions that can be reached by these transformations are relevant to various synchronous or asynchronous computing models.

Loop optimization plays an important role in compiler optimization and program transformation. Many sophisticated techniques such as loop-invariance code motion have been developed. Loop peeling is a technique to assist parallelization of loops by unfolding loops a few times. This paper introduces a novel technique called loop peeling based on quasi-invariance/induction variables. It aims at finding a general and automatic method to derive how many times a given loop should be peeled. Our technique allows for a number of iterations before some variables assigned inside a given loop become invariance or induction variables. In this paper we define the notion of quasi-invariance/induction variables, present an algorithm for statically computing the optimal peeling length of a given loop. Our technique can increase the accuracy of program analyses, improve the effectiveness of loop peeling and is well-suited as supporting other optimization techniques in the context of supercomputers.

Execution environments are used as specifications for specialization of input-output programs in the derivation of product lines. These environments, formalized as color-blind I/O-alternating transition systems, are tolerant to mutations in a given program's outputs. Execution environments enable new compiler optimizations, vastly exceeding usual reductions. We propose a notion of context-dependent refinement for I/O-alternating transition systems, which supports composition and hierarchical reuse. The framework is demonstrated by discussing adaptations to realistic design languages and by presenting an exampleof a product line.

The area of secure compilation aims to design compilers which produce hardened code that can withstand attacks from low-level co-linked components. So far, there is no formal correctness criterion for secure compilers that comes with a clear understanding of what security properties the criterion actually provides. Ideally, we would like a criterion that, if fulfilled by a compiler, guarantees that large classes of security properties of source language programs continue to hold in the compiled program, even as the compiled program is run against adversaries with low-level attack capabilities. This paper provides such a novel correctness criterion for secure compilers, called trace-preserving compilation (TPC). We show that TPC preserves a large class of security properties, namely all safety hyperproperties. Further, we show that TPC preserves more properties than full abstraction, the de-facto criterion used for secure compilation. Then, we show that several fully abstract compilers described in literature satisfy an additional, common property, which implies that they also satisfy TPC. As an illustration, we prove that a fully abstract compiler from a typed source language to an untyped target language satisfies TPC.

The functional correctness of compiler optimization does not ensure the security properties of the source program. A functionally correct compiler optimization may introduce new security vulnerabilities in the optimized program. In a compiler, ensuring the security of every optimization is a challenging task as it applies hundreds of optimizations. Thus, this work aims to develop a translation validation of information leakage for the optimization phase of the compiler without considering any intermediate information from the compiler. Our method measures the information leakage in a path, uses a concept of leak propagation over paths and loops, and defines the relative security between the source and optimized programs. This work considers two attack models based on different observation points to access the local memory and proposes different translation validation approaches. The correctness and complexity of the translation validation method are also provided. The experimental results in the SPARK compiler on various benchmarks show that the optimized program is not relatively secure in many cases.

We construct secret-key encryption SKE schemes that are secure against related-key attacks and in the presence of key-dependent messages RKA-KDM secure. We emphasize that RKA-KDM security is not merely the conjunction of individual security properties, but covers attacks in which ciphertexts of key-dependent messages under related keys are available. Besides being interesting in their own right, RKA-KDM secure schemes allow to garble circuits with XORs very efficiently Applebaum, TCC 2013. Until now, the only known RKA-KDM secure SKE scheme due to Applebaum is based on the LPN assumption. Our schemes are based on various other computational assumptions, namely DDH, LWE, QR, and DCR. We abstract from Applebaum's construction and proof, and formalize three generic technical properties that imply RKA-KDM security: one property is IND-CPA security, and the other two are the existence of suitable oracles that produce ciphertexts under related keys, resp.i¾?of key-dependent messages. We then give simple SKE schemes that achieve these properties. Our constructions are variants of known KDM-secure public-key encryption schemes. To additionally achieve RKA security, we isolate suitable homomorphic properties of the underlying schemes in order to simulate ciphertexts under related keys in the security proof. RKA-KDM security for our schemes holds w.r.t. affine functions over the respective mathematical domain. From a conceptual point of view, our work provides a generic and extensible way to construct encryption schemes with multiple special security properties.

Large software systems are often built from loosely-coupled subsystems. When a programmer uses a third-party software component as a building block of her system, she doesn’t want the code she imports to break the whole system. She needs some methods guaranteeing that linking the foreign software component to her system is safe. The most widely used methods for ensuring safe linking are type checking and code signing. Checking the type of the interfaces between two software components ensures that two components agree on the types they are using. Although type checking is quite strong and easy to use, it doesn’t guarantee that the code will behave in an expected way. Different static checking mechanisms have been suggested to address specific security properties of programs: a security-sensitive type system [5], wrappers which encapsulate untrusted programs and implement security-concerned properties [7], and so on. They give the users better facilities to address security properties than typical type-checking does, but they still suffer from a lack of expressiveness since their security or linking policies are fixed and encoded in their type or logic systems.

The increasing complexity and connectivity of modern embedded systems highlight the importance of runtime monitoring to ensure correctness and security. This poses a significant challenge, since monitoring tools can break extra-functional requirements such as timing constraints. Non-intrusive program tracing through side-channel analysis techniques have recently appeared in the literature and constitute a promising approach. Existing techniques, however, exhibit important limitations. In this paper, we present a novel technique for non-intrusive program tracing from power consumption, based on a signals and system analysis approach: we view the power consumption signal as the output of a system with the power consumption of training samples as input. Using spectral analysis, we compute the impulse response to identify the system; the intuition is that for the correct training sample, the system will appear close to a system that outputs a shifted copy of the input signal, for which the impulse response is an impulse at the position corresponding to the shift. We also use the Control Flow Graph from the source code to constrain the classifier to valid sequences only, leading to substantial performance improvements over previous works. Experimental results confirm the effectiveness of our technique and show its applicability to runtime monitoring. The experiments include tracing programs that execute randomly generated sequences of functions as well as tracing a real application developed with SCADE. The experimental evaluation also includes a case-study as evidence of the usability of our technique to detect anomalous execution through runtime monitoring.

An autonomous agent is one that is not only directed by its environment, but is also driven by internal motivation to achieve certain goals based on beliefs about the environmental behaviour. Design paradigms for autonomous agents such as belief-desire-intention take into account the agent’s “mental” features when presenting its patterns of behaviour. In this paper we present an approach to modelling autonomous agents by introducing mental features to conventional transition system specifications. Mental features such as belief and desire are represented by declarative linear temporal logic formulas. Refinement is then proposed to define the correctness of the agent design and development. It turns out, however, that the introduction of these mental features is not monotonic with respect to refinement. We therefore introduce additional refinement proof obligations to enable the use of simulation rules when checking refinement.

Data-parallel architectures must provide efficient support for complex control-flow constructs to support sophisticated applications coded in modern single-program multiple-data languages. As these architectures have wide data paths that process a single instruction across parallel threads, a mechanism is needed to track and sequence threads as they traverse potentially divergent control paths through the program. The design space for divergence management ranges from software-only approaches where divergence is explicitly managed by the compiler, to hardware solutions where divergence is managed implicitly by the micro architecture. In this paper, we explore this space and propose a new predication-based approach for handling control-flow structures in data-parallel architectures. Unlike prior predication algorithms, our new compiler analyses and hardware instructions consider the commonality of predication conditions across threads to improve efficiency. We prototype our algorithms in a production compiler and evaluate the tradeoffs between software and hardware divergence management on current GPU silicon. We show that our compiler algorithms make a predication-only architecture competitive in performance to one with hardware support for tracking divergence.

We present a denotational semantics for weak memory concurrency that avoids thin-air reads, provides data-race free programs with sequentially consistent semantics (DRF-SC), and supports a compositional refinement relation for validating optimisations. Our semantics identifies false program dependencies that might be removed by compiler optimisation, and leaves in place just the dependencies necessary to rule out thin-air reads. We show that our dependency calculation can be used to rule out thin-air reads in any axiomatic concurrency model, in particular C++. We present a tool that automatically evaluates litmus tests, show that we can augment C++ to fix the thin-air problem, and we prove that our augmentation is compatible with the previously used compilation mappings over key processor architectures. We argue that our dependency calculation offers a practical route to fixing the longstanding problem of thin-air reads in the C++ specification.

Compiler optimizations are critical to the efficiency of modern functional programs. At the same time, optimizations that unintentionally change the semantics of programs can systematically introduce errors into programs that pass through them. The question of how to best verify that optimizations and other program transformations preserve semantics is an important one, given the potential for error introduction. Dependent types allow us to prove that properties about our programs are correct, as well as to design data types and interpreters in such a way that they are correct-by-construction. In this thesis, we explore the use of dependent types and intrinsically-typed definitional interpreters in progressively larger subsets of Monadic Intermediate Language (MIL) to verify optimizations used in a compiler back end. In particular, we prove non-trivial program optimizations using the Agda proof assistant, and illustrate the benefits and challenges of this style of program verification.

Program development often proceeds by transforming simple, clear programs into complex, involuted, but more efficient ones. This paper examines ways this process can be rendered more systematic. We show how analysis of program performance, partial evaluation of functions, and abstraction of recursive function definitions from recurring subgoals can be combined to yield many global transformations in a methodical fashion. Examples are drawn from compiler optimization, list processing, very high-evel languages, and APL execution.