The widespread adoption of AI-assisted coding is directly proportional to an increase in software bugs; can AI-assisted formal verification help reduce bugs at a comparable scale? In this experience report we give an anecdotal account of AI agents, equipped with a CLI and a proof assistant, producing thousands of lines of machine-checked code. We detail our experience across different proof-engineering tasks: implementing verified data structures for a standard library, translating unverified code into a formal language while inferring its specification, and porting and refactoring existing proofs to new frameworks. We present the techniques that made agentic proof-oriented programming (PoP) effective---or ineffective---and characterize the role of the human expert, whose contribution reduces to providing natural-language problem descriptions, reviewing auto-generated specifications, and occasionally supplying a key invariant. Our findings suggest that this division of labor provides substantial leverage to the human expert in the loop: three experts, over the course of two weeks, completed case studies whose manual proof-engineering cost we estimate at roughly half a year.

We introduce Kuiper, a language for safe and verified efficient CPU/GPU programming embedded as an extensible library within the F* dependently typed language. We rely on F*’s support for dependent types and its associated Pulse concurrent separation logic to develop a program logic in which to prove CPU/GPU programs safe, data-race free, and functionally correct. Our model of the GPU includes several intricacies, including the memory hierarchy, kernel launches, and synchronization within a single comprehensive framework. To do so, we extend the Pulse program logic with a novel notion of located resources and a new connective to structure reasoning about massively parallel programs, and present new proof rules to lift the per-thread view of GPU kernels to an end-to-end correctness specification. We have used Kuiper to program and prove correct a variety of GPU kernels, including full functional correctness proofs of an optimized matrix multiplication using two levels of block tiling and tensor cores. In doing so, we have developed a range of libraries to enable programs and proofs at a high level of abstraction but without imposing any runtime overhead. These allow Kuiper programs to be polymorphic (over types, operations, memory layout, and more) and compile to efficient, specialized CUDA code, while enabling a novel form of verified auto-tuning. Our experimental evaluation confirms that Kuiper programs match the performance of their handwritten CUDA counterparts and are competitive with closed source, state-of-the-art kernels in cuBLAS.

Project Everest began at Microsoft Research in 2016, aiming to spur research in program verification to produce industrial-grade software. In collaboration with INRIA and Carnegie Mellon University, Project Everest’s goal was to produce drop-in verified replacements of secure communications software used in the HTTPS ecosystem, including TLS, the underlying cryptography, and related subprotocols. Now, almost a decade later, we reflect on the project, sharing both its successes and failures, and look ahead to the next decade of program verification research.

Incorrect handling of security-critical data formats, particularly in low-level languages, are the root cause of many security vulnerabilities. Provably correct parsing and serialization tools that target languages like C can help. Towards this end, we present PulseParse, a library of verified parser and serializer combinators for non-malleable binary formats. Specifications and proofs in PulseParse are in separation logic, offering a more abstract and compositional interface, with full support for data validation, parsing, and serialization. PulseParse also supports a class of recursive formats---with a focus on security and handling adversarial inputs, we show how to parse such formats with only a constant amount of stack space.

Improper parsing of attacker-controlled input is a leading source of software security vulnerabilities, especially when programmers transcribe informal format descriptions in RFCs into efficient parsing logic in low-level, memory unsafe languages. Several researchers have proposed formal specification languages for data formats from which efficient code can be extracted. However, distilling informal requirements into formal specifications is challenging and, despite their benefits, new, formal languages are hard for people to learn and use. In this work, we present 3DGen, a framework that makes use of AI agents to transform mixed informal input, including natural language documents (i.e., RFCs) and example inputs into format specifications in a language called 3D. To support humans in understanding and trusting the generated specifications, 3DGen uses symbolic methods to also synthesize test inputs that can be validated against an external oracle. Symbolic test generation also helps in distinguishing multiple plausible solutions. Through a process of repeated refinement, 3DGen produces a 3D specification that conforms to a test suite, and which yields safe, efficient, provably correct, parsing code in C. We have evaluated 3DGen on 20 Internet standard formats, demonstrating the potential for AI-agents to produce formally verified$C$code at a non-trivial scale. A key enabler is the use of a domain-specific language to limit AI outputs to a class for which automated. symbolic analysis is tractable.

PulseCore is a new program logic suitable for intrinsic proofs of higher-order, stateful, concurrent, dependently typed programs. It provides many of the features of a modern, concurrent separation logic, including dynamically allocated impredicative invariants, higher-order ghost state, step-indexing with later credits, and support for user-defined ghost state constructions. PulseCore is developed foundationally within the F ⋆ programming language with fully mechanized proofs, and is applicable to F ⋆ programs itself. To evaluate our work, we use Pulse , a surface language within F ⋆ for PulseCore , to develop a range of program proofs. Illustrating its suitability for proving higher-order concurrent programs, we present a verified library for task pools in the style of OCaml5, together with some verified task-parallel programs. Next, we present various data structures and synchronization primitives, including a barrier that requires the use of higher-order ghost state. Finally, we present a verified implementation of the DICE Protection Environment, an industry standard secure boot protocol. Taken together, our evaluation consists of more than 31,000 lines of verified code in a range of settings, providing evidence that PulseCore is both highly expressive as well as practical for a variety of program proof applications.

FastVer is a protocol that uses a variety of memory-checking techniques to monitor the integrity of key-value stores with only a modest runtime cost. Arasu et al. formalize the high-level design of FastVer in the F* proof assistant and prove it correct. However, their formalization did not yield a provably correct implementation---FastVer is implemented in unverified C++ code.

Abstract Syntax Notation One (ASN.1) is a language for structured data exchange between computers, standardized by both ITU-T and ISO/IEC since 1984. The Distinguished Encoding Rules (DER) specify its non-malleable binary format: for a given ASN.1 data type, every value has a distinct, unique binary representation. ASN.1 DER is used in many security-critical interfaces for telecommunications and networking, such as the X.509 public key infrastructure, where non-malleability is essential. However, due to the expressiveness and flexibility of the general-purpose ASN.1 language, correctly parsing ASN.1 DER data formats is still considered a serious security challenge in practice. We present ASN1*, the first formalization of ASN.1 DER with a mechanized proof of non-malleability. Our development provides a shallow embedding of ASN.1 in the F* proof assistant and formalizes its DER semantics within the EverParse parser generator framework. It guarantees that any ASN.1 data encoded using our DER semantics is non-malleable. It yields verified code that parses valid binary representations into values of the corresponding ASN.1 data type while rejecting invalid ones. We empirically confirm that our semantics models ASN.1 DER usage in practice by evaluating ASN1* parsers extracted to OCaml on both positive and negative test cases involving X.509 certificates and Certificate Revocation Lists (CRLs).

With an eye toward performance, interoperability, or legacy concerns, low-level system software often must parse binary encoded data formats. Few tools are available for this task, especially since the formats involve a mixture of arithmetic and data dependence, beyond what can be handled by typical parser generators. As such, parsers are written by hand in languages like C, with inevitable errors leading to security vulnerabilities.

We present FastVer, a high-performance key-value store with strong data integrity guarantees. FastVer is built as an extension of FASTER, an open-source, high-performance key-value store. It offers the same key-value API as FASTER plus an additional verify() method that detects if an unauthorized attacker tampered with the database and checks whether results of all read operations are consistent with historical updates. FastVer is based on a novel approach that combines the advantages of Merkle trees and deferred memory verification. We show that this approach achieves one to two orders of magnitudes higher throughputs than traditional approaches based on either Merkle trees or memory verification. We have formally proven the correctness of our approach in a proof assistant, ensuring that verify() detects any inconsistencies, except if a collision can be found on a cryptographic hash.

Drawing on earlier protocol-verification work, we investigate the security of the QUIC record layer, as standardized by the IETF in draft version 30. This version features major differences compared to Google’s original protocol and early IETF drafts. It serves as a useful test case for our verification methodology and toolchain, while also, hopefully, drawing attention to a little studied yet crucially important emerging standard.We model QUIC packet and header encryption, which uses a custom construction for privacy. To capture its goals, we propose a security definition for authenticated encryption with semi-implicit nonces. We show that QUIC uses an instance of a generic construction parameterized by a standard AEAD-secure scheme and a PRF-secure cipher. We formalize and verify the security of this construction in F. The proof uncovers interesting limitations of nonce confidentiality, due to the malleability of short headers and the ability to choose the number of least significant bits included in the packet counter. We propose improvements that simplify the proof and increase robustness against strong attacker models. In addition to the verified security model, we also give a concrete functional specification for the record layer, and prove that it satisfies important functionality properties (such as the correct successful decryption of encrypted packets) after fixing more errors in the draft. We then provide a high-performance implementation of the record layer that we prove to be memory safe, correct with respect to our concrete specification (inheriting its functional correctness properties), and secure with respect to our verified model. To evaluate this component, we develop a provably-safe implementation of the rest of the QUIC protocol. Our record layer achieves nearly 2 GB/s throughput, and our QUIC implementation’s performance is within 21% of an unverified baseline.

Steel is a language for developing and proving concurrent programs embedded in F ⋆ , a dependently typed programming language and proof assistant. Based on SteelCore, a concurrent separation logic (CSL) formalized in F ⋆ , our work focuses on exposing the proof rules of the logic in a form that enables programs and proofs to be effectively co-developed. Our main contributions include a new formulation of a Hoare logic of quintuples involving both separation logic and first-order logic, enabling efficient verification condition (VC) generation and proof discharge using a combination of tactics and SMT solving. We relate the VCs produced by our quintuple system to solving a system of associativity-commutativity (AC) unification constraints and develop tactics to (partially) solve these constraints using AC-matching modulo SMT-dischargeable equations. Our system is fully mechanized and implemented in F ⋆ . We evaluate it by developing several verified programs and libraries, including various sequential and concurrent linked data structures, proof libraries, and a library for 2-party session types. Our experience leads us to conclude that our system enables a mixture of automated and interactive proof, making it productive to build programs foundationally verified against a highly expressive, state-of-the-art CSL.

We present EverCrypt: a comprehensive collection of verified, high-performance cryptographic functionalities available via a carefully designed API. The API provably supports agility (choosing between multiple algorithms for the same functionality) and multiplexing (choosing between multiple implementations of the same algorithm). Through abstraction and zero-cost generic programming, we show how agility can simplify verification without sacrificing performance, and we demonstrate how C and assembly can be composed and verified against shared specifications. We substantiate the effectiveness of these techniques with new verified implementations (including hashes, Curve25519, and AES-GCM) whose performance matches or exceeds the best unverified implementations. We validate the API design with two high-performance verified case studies built atop EverCrypt, resulting in line-rate performance for a secure network protocol and a Merkle-tree library, used in a production blockchain, that supports 2.7 million insertions/sec. Altogether, EverCrypt consists of over 124K verified lines of specs, code, and proofs, and it produces over 29K lines of C and 14K lines of assembly code.

We introduce Meta-F $$^{\star }$$ , a tactics and metaprogramming framework for the F $$^\star $$ program verifier. The main novelty of Meta-F $$^\star $$ is allowing the use of tactics and metaprogramming to discharge assertions not solvable by SMT, or to just simplify them into well-behaved SMT fragments. Plus, Meta-F $$^\star $$ can be used to generate verified code automatically. Meta-F $$^\star $$ is implemented as an F $$^\star $$ effect, which, given the powerful effect system of F $$^{\star }$$ , heavily increases code reuse and even enables the lightweight verification of metaprograms. Metaprograms can be either interpreted, or compiled to efficient native code that can be dynamically loaded into the F $$^\star $$ type-checker and can interoperate with interpreted code. Evaluation on realistic case studies shows that Meta-F $$^\star $$ provides substantial gains in proof development, efficiency, and robustness.

Relational properties describe multiple runs of one or more programs. They characterize many useful notions of security, program refinement, and equivalence for programs with diverse computational effects, and they have received much attention in the recent literature. Rather than developing separate tools for special classes of effects and relational properties, we advocate using a general purpose proof assistant as a unifying framework for the relational verification of effectful programs. The essence of our approach is to model effectful computations using monads and to prove relational properties on their monadic representations, making the most of existing support for reasoning about pure programs. We apply this method in F* and evaluate it by encoding a variety of relational program analyses, including information flow control, program equivalence and refinement at higher order, correctness of program optimizations and game-based cryptographic security. By relying on SMT-based automation, unary weakest preconditions, user-defined effects, and monadic reification, we show that, compared to unary properties, verifying relational properties requires little additional effort from the F* programmer.

Concurrent abstraction layers are ubiquitous in modern computer systems because of the pervasiveness of multithreaded programming and multicore hardware. Abstraction layers are used to hide the implementation details (e.g., fine-grained synchronization) and reduce the complex dependencies among components at different levels of abstraction. Despite their obvious importance, concurrent abstraction layers have not been treated formally. This severely limits the applicability of layer-based techniques and makes it difficult to scale verification across multiple concurrent layers.

We present Low*, a language for low-level programming and verification, and its application to high-assurance optimized cryptographic libraries. Low* is a shallow embedding of a small, sequential, well-behaved subset of C in F*, a dependently- typed variant of ML aimed at program verification. Departing from ML, Low* does not involve any garbage collection or implicit heap allocation; instead, it has a structured memory model à la CompCert, and it provides the control required for writing efficient low-level security-critical code. By virtue of typing, any Low* program is memory safe. In addition, the programmer can make full use of the verification power of F* to write high-level specifications and verify the functional correctness of Low* code using a combination of SMT automation and sophisticated manual proofs. At extraction time, specifications and proofs are erased, and the remaining code enjoys a predictable translation to C. We prove that this translation preserves semantics and side-channel resistance. We provide a new compiler back-end from Low* to C and, to evaluate our approach, we implement and verify various cryptographic algorithms, constructions, and tools for a total of about 28,000 lines of code. We show that our Low* code delivers performance competitive with existing (unverified) C cryptographic libraries, suggesting our approach may be applicable to larger-scale low-level software.

We provide concrete evidence that floating-point computations in C programs can be verified in a homogeneous verification setting based on Coq only, by evaluating the practicality of the combination of the formal semantics of CompCert Clight and the Flocq formal specification of IEEE 754 floating-point arithmetic for the verification of properties of floating-point computations in C programs. To this end, we develop a framework to automatically compute real-number expressions of C floating-point computations with rounding error terms along with their correctness proofs. We apply our framework to the complete analysis of an energy-efficient C implementation of a radar image processing algorithm, for which we provide a certified bound on the total noise introduced by floating-point rounding errors and energy-efficient approximations of square root and sine.

Recent ground-breaking efforts such as CompCert have made a convincing case that mechanized verification of the compiler correctness for realistic C programs is both viable and practical. Unfortunately, existing verified compilers can only handle whole programs---this severely limits their applicability and prevents the linking of verified C programs with verified external libraries. In this paper, we present a novel compositional semantics for reasoning about open modules and for supporting verified separate compilation and linking. More specifically, we replace external function calls with explicit events in the behavioral semantics. We then develop a verified linking operator that makes lazy substitutions on (potentially reacting) behaviors by replacing each external function call event with a behavior simulating the requested function. Finally, we show how our new semantics can be applied to build a refinement infrastructure that supports both vertical composition and horizontal composition.

Modern computer systems consist of a multitude of abstraction layers (e.g., OS kernels, hypervisors, device drivers, network protocols), each of which defines an interface that hides the implementation details of a particular set of functionality. Client programs built on top of each layer can be understood solely based on the interface, independent of the layer implementation. Despite their obvious importance, abstraction layers have mostly been treated as a system concept; they have almost never been formally specified or verified. This makes it difficult to establish strong correctness properties, and to scale program verification across multiple layers.

Verified compilers guarantee the preservation of semantic properties and thus enable formal verification of programs at the source level. However, important quantitative properties such as memory and time usage still have to be verified at the machine level where interactive proofs tend to be more tedious and automation is more challenging.

We present a formal operational semantics and its Coq mechanization for the C++ object model, featuring object construction and destruction, shared and repeated multiple inheritance, and virtual function call dispatch. These are key C++ language features for high-level system programming, in particular for predictable and reliable resource management. This paper is the first to present a formal mechanized account of the metatheory of construction and destruction in C++, and applications to popular programming techniques such as "resource acquisition is initialization". We also report on irregularities and apparent contradictions in the ISO C++03 and C++11 standards.

Object layout - the concrete in-memory representation of objects - raises many delicate issues in the case of the C++ language, owing in particular to multiple inheritance, C compatibility and separate compilation. This paper formalizes a family of C++ object layout schemes and mechanically proves their correctness against the operational semantics for multiple inheritance of Wasserrab et al. This formalization is flexible enough to account for space-saving techniques such as empty base class optimization and tail-padding optimization. As an application, we obtain the first formal correctness proofs for realistic, optimized object layout algorithms, including one based on the popular "common vendor" Itanium C++ application binary interface. This work provides semantic foundations to discover and justify new layout optimizations; it is also a first step towards the verification of a C++ compiler front-end.

Abstract This paper explains how the Alloy model-finding method has been used to check the specification of an electronic purse (also called smart card) system, called the Mondex case study, initially written in Z. After describing the payment protocol between two electronic purses, and presenting an overview of the Alloy model-finding method, this paper explains how technical issues about integers and conceptual issues about the object layout in Z have been tackled in Alloy, giving general methods that can be used in most case studies with Alloy. This work has also pointed out some significant bugs in the original Z specification such as reasoning bugs in the proofs, and proposes a way to solve them.