eBPF is a technology that allows developers to safely extend kernel functionality without modifying kernel source code or developing loadable kernel modules. Since the kernel governs critical system operations and enforces isolation boundaries between user space and privileged data, any mechanism that modifies its behavior must meet the highest standards of safety and correctness. To this end, the eBPF toolchain includes a verifier, which statically checks safety properties such as memory access validity, bounded loops, and type correctness before loading the program into the kernel. However, the existing verifier is both overly conservative in some cases-rejecting valid programs-and unsound in others, permitting unsafe behavior that violates the intended semantics of the kernel interface. To address these challenges, we introduce BeePL, a domain-specific language for eBPF with a formally verified type system. The BeePL type system, along with the language design, statically enforces key safety properties such as type-correct memory access, safe pointer usage, absence of unbounded loops, and structured control flow. These guarantees are backed by formal type soundness proofs, ensuring that well-typed programs satisfy the safety invariants required by the eBPF execution environment. BeePL also proves that well-typed source programs meet critical eBPF-specific properties related to memory safety, termination, and control flow, enabling high-level reasoning prior to compilation. For properties not fully enforceable statically-such as dynamic bounds and undefined behavior-BeePL inserts semantics-preserving runtime checks during compilation. We develop a verified compilation strategy that extends CompCert to generate BPF bytecode from BeePL programs, establishing a principled foundation for an end-to-end verifiable toolchain for safe kernel extensions.

Modern operating systems have adopted Berkeley Packet Filters (BPF) as a mechanism to extend kernel functionalities dynamically, e.g., Linux's eBPF or RIOT's rBPF. The just-in-time (JIT) compilation of eBPF introduced in Linux eBPF for performance has however led to numerous critical issues. Instead, RIOT's rBPF uses a slower but memory-isolating interpreter (a virtual machine) which implements a defensive semantics of BPF; and therefore trades performance for security. To increase performance without sacrificing security, this paper presents a fully verified JIT implementation for RIOT's rBPF, consisting of: i/ an end-to-end refinement workflow to both proving the JIT correct from an abstract specification and by deriving a verified concrete C implementation; ii/ a symbolic CompCert interpreter for executing jited binary code; iii/ a verified JIT compiler for rBPF; iv/ a verified hybrid rBPF virtual machine. Our core contribution is, to the best of our knowledge, the first and fully verified rBPF JIT compiler with correctness guarantees from high-level specification to low-level implementation. Benchmarks on microcontrollers hosting the RIOT operating system demonstrate significant performance improvements over the existing implementations of rBPF, even in worst-case application scenarios.

We present PfComp, a verified compiler for stateless firewall policies. The policy is first compiled into an intermediate representation taking the form of a binary decision diagram that is optimised in terms of decision nodes. The decision diagram is then compiled into a Clight program. The compiler is proved correct using the Coq proof assistant and extracted into OCaml code. Our preliminary experiments show promising results. The compiler generates code for relatively large firewall policies and the generated code out-performs a sequential evaluation of the policy rules.

Constant-time programming is the de facto standard to protect security-sensitive software against cache-based timing attacks. This software countermeasure is effective but may incur a significant performance overhead and require a substantial rewrite of the code. In this work, we study a secure cache-locking hardware mechanism which eases the writing of secure code and has little execution overhead. To reason about the security of software, we propose a high-level leakage model such that accesses to locked memory addresses do not generate any observable leakage. To ensure the adequacy of this leakage model, we also propose a concrete hardware leakage model for a RISC-V microcontroller where the secure code may be interrupted, at any time, by some arbitrary malicious code. Using the Observational Non-Interference setting, we show formally that the security of the software model is preserved at the hardware level. We evaluate the effectiveness and performance of this mechanism, notably on block ciphers. We also propose and evaluate a new constant-time sorting algorithm.

Timing side-channels are an identified threat for security critical software. Existing countermeasures have a cost either on the hardware requirements or execution time. We focus on low-cost microcontrollers that have a very low computational capacity. Although these processors do not feature out-of-order execution or speculation, they remain vulnerable to timing attacks exploiting the varying latencies of ALU operations or memory accesses.We propose to augment the RISC-V ISA with security primitives that have a guaranteed timing behavior. These primitives allow constant time ALU operations and memory accesses that do not alter the state of the cache. Our approach has a low overhead in terms of hardware cost, binary code size, and execution time both for the constant time secure program and other programs running concurrently on the same hardware.

As a revolutionary kernel extension technology, Berkeley Packet Filters (BPF) has been applied for various operating systems from different domains, from servers (Linux’s extended BPF) to microcontrollers (RIOT-OS rBPF). Previous works have formally proved the memory isolation property for the non-optimized rBPF virtual machine in the Coq proof assistant. In this paper, we introduce a verified optimization for rBPF, and highlight a novel proof approach for optimization correctness: a simplification process is first used to transform monadic models with option state to simplified non-monadic models with inline arguments; then the optimization correctness theorem is split into i) proving simplification correct and ii) proving the optimization correctness on simplified models. Our proof approach enjoys a fruitful proof simplification. Preliminary experiments demonstrate satisfying performance.

Constant-time is a programming discipline which protects security sensitive code against a wide class of timing attacks. This discipline can be formalised as a non-interference property and enforced by an information flow type system which prevents branching and memory accesses over secret data. We propose a relaxed information flow type system which tracks indirect flows but only rejects programs leaking secrets through direct flows. The main result of this paper is that any program that is accepted using this relaxed type system can be transformed automatically into a semantically equivalent constant-time program. Our algorithms are implemented in the jasmin compiler and validated against synthetic programs.

Low-power operating system runtimes used on IoT microcontrollers typically provide rudimentary APIs, basic connectivity and, sometimes, a (secure) firmware update mechanism. In contrast, on less constrained hardware, networked software has entered the age of serverless, microservices and agility. With a view to bridge this gap, in the paper we design Femto-Containers, a new middleware runtime which can be embedded on heterogeneous low-power IoT devices. Femto-Containers enable the secure deployment, execution and isolation of small virtual software functions on low-power IoT devices, over the network. We implement Femto-Containers, and provide integration in RIOT, a popular open source IoT operating system. We then evaluate the performance of our implementation, which was formally verified for fault-isolation, guaranteeing that RIOT is shielded from logic loaded and executed in a Femto-Container. Our experiments on various popular microcontroller architectures (Arm Cortex-M, ESP32 and RISC-V) show that Femto-Containers offer an attractive trade-off in terms of memory footprint overhead, energy consumption, and security

RIOT is a micro-kernel dedicated to IoT applications that adopts eBPF (extended Berkeley Packet Filters) to implement so-called femto-containers. As micro-controllers rarely feature hardware memory protection, the isolation of eBPF virtual machines (VM) is critical to ensure system integrity against potentially malicious programs. This paper shows how to directly derive, within the Coq proof assistant, the verified C implementation of an eBPF virtual machine from a Gallina specification. Leveraging the formal semantics of the CompCert C compiler, we obtain an end-to-end theorem stating that the C code of our VM inherits the safety and security properties of the Gallina specification. Our refinement methodology ensures that the isolation property of the specification holds in the verified C implementation. Preliminary experiments demonstrate satisfying performance.

We present the design and implementation of itauto , a Coq reflexive tactic for intuitionistic propositional logic. The tactic inherits features found in modern SAT solvers: definitional conjunctive normal form; lazy unit propagation and conflict driven backjumping. Formulae are hash-consed using native integers thus enabling a fast equality test and a pervasive use of Patricia Trees. We also propose a hybrid proof by reflection scheme whereby the extracted solver calls user-defined tactics on the leaves of the propositional proof search thus enabling theory reasoning and the generation of conflict clauses. The solver has decent efficiency and is more scalable than existing tactics on synthetic benchmarks and preliminary experiments are encouraging for existing developments.

Software Fault Isolation (SFI) is a security-enhancing program transformation for instrumenting an untrusted binary module so that it runs inside a dedicated isolated address space, called a sandbox. To ensure that the untrusted module cannot escape its sandbox, existing approaches such as Google's Native Client rely on a binary verifier to check that all memory accesses are within the sandbox. Instead of relying on a posteriori verification, we design, implement and prove correct a program instrumentation phase as part of the formally verified compiler CompCert that enforces a sandboxing security property a priori. This eliminates the need for a binary verifier and, instead, leverages the soundness proof of the compiler to prove the security of the sandboxing transformation. The technical contributions are a novel sandboxing transformation that has a well-defined C semantics and which supports arbitrary function pointers, and a formally verified C compiler that implements SFI. Experiments show that our formally verified technique is a competitive way of implementing SFI.

Correct compilers perform program transformations preserving input/output behaviours of programs. Yet, correctness does not prevent program optimisations from introducing information-flow leaks that would make the target program more vulnerable to side-channel attacks than the source program. To tackle this problem, we propose a notion of Information-Flow Preserving (IFP) program transformation which ensures that a target program is no more vulnerable to passive side-channel attacks than a source program. To protect against a wide range of attacks, we model an attacker who is granted arbitrary memory accesses for a pre-defined set of observation points. We propose a compositional proof principle for proving that a transformation is IFP. Using this principle, we show how a translation validation technique can be used to automatically verify and even close information-flow leaks introduced by standard compiler passes such as dead-store elimination and register allocation. The technique has been experimentally validated on the CompCert C compiler.

Correct compilers perform program transformations preserving input/output behaviours of programs. Yet, correctness does not prevent program optimisations from introducing information-flow leaks that would make the target program more vulnerable to side-channel attacks than the source program. To tackle this problem, we propose a notion of Information-Flow Preserving (IFP) program transformation which ensures that a target program is no more vulnerable to passive side-channel attacks than a source program. To protect against a wide range of attacks, we model an attacker who is granted arbitrary memory accesses for a pre-defined set of observation points. We propose a compositional proof principle for proving that a transformation is IFP. Using this principle, we show how a translation validation technique can be used to automatically verify and even close information-flow leaks introduced by standard compiler passes such as dead-store elimination and register allocation. The technique has been experimentally validated on the CompCert C compiler.

Software Fault Isolation (SFI) consists in transforming untrusted code so that it runs within a specific address space, (called the sandbox) and verifying at load-time that the binary code does indeed stay inside the sandbox. Security is guaranteed solely by the SFI verifier whose correctness therefore becomes crucial. Existing verifiers enforce a very rigid, almost syntactic policy where every memory access and every control-flow transfer must be preceded by a sandboxing instruction sequence, and where calls outside the sandbox must implement a sophisticated protocol based on a shadow stack. We propose to define SFI as a defensive semantics, with the purpose of deriving semantically sound verifiers that admit flexible and efficient implementations of SFI. We derive an executable analyser, that works on a per-function basis, which ensures that the defensive semantics does not go wrong, and hence that the code is well isolated. Experiments show that our analyser exhibits the desired flexibility: it validates correctly sandboxed code, it catches code breaking the SFI policy, and it can validate programs where redundant instrumentations are optimised away.

A common security recommendation is to reduce the in-memory lifetime of secret values, in order to reduce the risk that an attacker can obtain secret data by probing memory. To mitigate this risk, secret values can be overwritten, at source level, after their last use. The problem we address here is how to ensure that a compiler preserve these mitigation efforts and thus that secret values are not easier to obtain at assembly level than at source level. We propose a formal definition of Information Flow Preserving program Transformations in which we model the information leak of a program using the notion of Attacker Knowledge. Program transformations are validated by relating the knowledge of the attacker before and after the transformation. We consider two classic compiler passes (Dead Store Elimination and Register Allocation) and show how to validate and, if needed, modify these transformations in order to be information flow preserving.

The CompCert C compiler provides the formal guarantee that the observable behaviour of the compiled code improves on the observable behaviour of the source code. In this paper, we present a formally verified C compiler, CompCertS, which is essentially the CompCert compiler, albeit with a stronger formal guarantee: it gives a semantics to more programs and ensures that the memory consumption is preserved by the compiler. CompCertS is based on an enhanced memory model where, unlike CompCert but like Gcc, the binary representation of pointers can be manipulated much like integers and where, unlike CompCert, allocation may fail if no memory is available. The whole proof of CompCertS is a significant proof-effort and we highlight the crux of the novel proofs of 12 passes of the back-end and a challenging proof of an essential optimising pass of the front-end.

The Coq proof-assistant provides automation for various logic fragments. However, there is a lack of generic support for extending those tactics. To augment the proof automation, we propose an extensible reflexive tactic, ppsimpl, aiming at canonising goals so that the scope of existing tactics can be broaden at low cost. The tactic first injects a type, say T, into a canonical type, say CT and maps function over T into their counterpart over CT. This transformation allows type T to benefit from the automation provided for type CT. The tactic also performs another normalisation step which purpose is to restrict the operators of CT to those that are known to the automated tactics. This is done by either unfolding function definitions or replacing a function by a (partial) specification. The extensibility of the ppsimpl tactic is obtained through the type-class mechanism which allows to infer and collect all the necessary proof objects. These instances are processed by a Ltac compiler which automatically generate the reification of terms and instantiate the generic correctness proof.

The CompCert C compiler guarantees that the target program behaves as the source program. Yet, source programs without a defined semantics do not benefit from this guarantee and could therefore be miscompiled. To reduce the possibility of a miscompilation, we propose a novel memory model for CompCert which gives a defined semantics to challenging features such as bitwise pointer arithmetics and access to uninitialised data. We evaluate our memory model both theoretically and experimentally. In our experiments, we identify pervasive low-level C idioms that require the additional expressiveness provided by our memory model. We also show that our memory model provably subsumes the existing CompCert memory model thus cross-validating both semantics. Our memory model relies on the core concepts of symbolic value and normalisation. A symbolic value models a delayed computation and the normalisation turns, when possible, a symbolic value into a genuine value. We show how to tame the expressive power of the normalisation so that the memory model fits the proof framework of CompCert. We also adapt the proofs of correctness of the compiler passes performed by CompCert's front-end, thus demonstrating that our model is well-suited for proving compiler transformations.