Formally Discovering and Reproducing Network Protocols Vulnerabilities
Christophe Crochet 0000-0001-8635-2098, John Aoga 0000-0002-7213-146X, Axel Legay 0000-0003-2287-8925 · 2025
Read Formally Discovering and Reproducing Network Protocols Vulnerabilities by Christophe Crochet 0000-0001-8635-2098, John Aoga 0000-0002-7213-146X, Axel Legay 0000-0003-2287-8925 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-01538.txt.
Introduction
Christophe Crochet, John Aoga, and Axel Legay (2025) present a formal method for finding security flaws in network protocols, and for reproducing them. Protocols grow ever more complex, and hand-written attacks miss corner cases. Their method, called NACT, builds attacks from composable pieces. Attacker specifications can be combined. The protocol's formal specification can be mutated. And randomized constraint solving searches for inputs that satisfy an attack's conditions. The pieces generate scenarios and test cases automatically.
The authors validate the approach on a custom protocol and on five used versions of QUIC, the protocol behind much modern web traffic. NACT found and reproduced real flaws, including abuse of version negotiation, the handshake step where endpoints agree what dialect to speak. Comparing current and older versions showed which flaws had been fixed and which persisted. The goal is protocol testing that proves an attack works, rather than merely suspecting it might.
Catalog record
| CREATOR | Christophe Crochet 0000-0001-8635-2098, John Aoga 0000-0002-7213-146X, Axel Legay 0000-0003-2287-8925 |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 41 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-01538
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Formally Discovering and Reproducing Network Protocols Vulnerabilities Christophe Crochet 0000-0001-8635-2098, John Aoga 0000-0002-7213-146X, Axel Legay 0000-0003-2287-8925 Abstract The rapid evolution of cyber threats has increased the need for robust methods to discover vulnerabilities in increasingly complex and diverse network protocols. This paper introduces Network Attack-centric Compositional Testing ( NACT ) [ 12 ] , a novel methodology designed to discover new vulnerabilities in network protocols and create scenarios to reproduce these vulnerabilities through attacker models. NACT integrates composable attacker specifications, formal specification mutations, and randomized constraint-solving techniques to generate sophisticated attack scenarios and test cases. The methodology enables comprehensive testing of both single-protocol and multi-protocol interactions. Through case studies involving a custom minimalist protocol ( MiniP ) and five widely used QUIC implementations, NACT is shown to effectively identify, reproduce, and find new real-world vulnerabilities such as version negotiation abuse. Additionally, by comparing the current and older versions of these QUIC implementations, NACT demonstrates its ability to detect both persistent vulnerabilities and regressions. Finally, by supporting cross-protocol testing within a black-box testing framework, NACT provides a versatile approach to improve the security of network protocols. 1 Introduction In today’s hyperconnected world, network protocols serve as the foundation of global communication infrastructures, yet they remain vulnerable to sophisticated cyber threats. Protocol vulnerabilities provide a fertile ground for attackers to disrupt services, steal sensitive information, and cause widespread damage, as seen in high-profile incidents such as the Heartbleed [ 37 ] and Log4Shell [ 15 ] attacks. The unpredictability of cyberattacks lies in their ability to exploit weaknesses in ways not anticipated during the protocol’s design. Attackers frequently leverage zero-day vulnerabilities and subtle deviations from formal specifications to create unexpected behaviors that lead to breaches [ 40 ] . This challenge highlights the necessity of anticipating attacks by uncovering potential vulnerabilities before they can be exploited [ 38 ] . Testing and simulation approaches play a crucial role in this proactive defense strategy. Previous studies have delved into various fault-based testing methodologies. Some methods focus on validating protocols according to their defined specifications (conformance testing) [ 32 , 16 ] , while others target known vulnerabilities directly [ 35 , 36 ] , especially through mutation testing [ 41 ] , which creates test scenarios by injecting minor controlled errors (mutants) into protocols to gauge their resilience [ 33 ] . These strategies typically fall under the umbrella of black-box or fuzz testing [ 36 ] . It is crucial to perform tests directly on implementations to enhance the realism and applicability of these tests for real-world network situations. This is where Model-Based Testing (MBT) [ 28 ] comes in, using an advanced language to describe protocol specifications and validate them directly against real implementations. This method facilitates the identification of potential code vulnerabilities that an attacker can exploit in practical scenarios, especially given the growing complexity of modern protocol implementations. Recently, Network-centric Compositional Testing ( NCT ) [ 23 ] has emerged as a pioneering MBT approach [ 28 ] that uses formal protocol specifications, implemented with tools such as Ivy [ 29 ] , to generate test tools automatically. NCT has effectively identified bugs and vulnerabilities, particularly in real-world protocols such as QUIC . However, NCT focused on individual compliance checks and did not address the dynamic and diverse nature of actual network interactions, which are essential to detect potential vulnerabilities before they can be exploited. To bridge this gap, we present Network Attack-centric Compositional Testing ( NACT ), a novel model-based mutation testing that builds upon NCT . NACT is tailored to thoroughly evaluate real-world network protocols for security flaws. Using compositional attacker models, mutations in formal specifications, and randomized constraint-solving techniques, NACT can uncover previously unknown vulnerabilities by emulating various advanced attack scenarios. Our method facilitates the examination of multi-protocol and cross-protocol interactions in black-box settings, thereby creating realistic, adversarial circumstances reflective of potential exploits in live environments. The capability to conduct multi-protocol testing and multipoint communications is becoming increasingly crucial with the rise of microservice architectures and distributed systems. We validate our method’s effectiveness by testing several recent vulnerabilities in the widely used QUIC protocol, successfully identifying new significant security weaknesses, including version negotiation issues and buffer overflows. The remainder of the paper is organized as follows. Section 2 offers an overview of network protocol vulnerabilities, associated attacks, the NCT methodology, and its inherent limitations. Section 3 introduces the NACT methodology, elaborating on the mutations of formal specification, their implementation, and integration within a compositional testing framework. Section 4 discusses our case studies on five QUIC implementations, showcasing the ability of NACT to identify complex vulnerabilities. Finally, Section 5 concludes and suggests possible directions for future work. 2 Background This section covers the core concepts of network attacks, how to test network protocols against these threats using network-centric approaches, and the limitations of these approaches, emphasizing the need for more comprehensive testing frameworks. (1) Network Attack Vectors and Agents. In network security, a vulnerability is a defect or weakness present in the design, implementation, or operation of a system that an attacker can exploit to breach the system’s security attributes, such as confidentiality , integrity , availability , and authenticity . When a malicious entity discovers and exploits a vulnerability, it results in a network attack , which is any intentional action to disrupt standard system functioning. (a) Types of Network Attacks. To understand the different forms of network attacks, we conducted a literature review and identified the most prevalent types of attacks that can occur in network environments. We classify these attacks according to their mode of operation. Table 1 summarizes the general categories of network attacks as referenced in various works [ 8 , 18 , 21 , 10 ] . Type of Attack Description TLS/SSL Connection Attacks Exploiting vulnerabilities in the TLS handshake process, such as protocol downgrade or renegotiation attacks, to intercept, decrypt, or alter encrypted communication data. Injection Injecting malicious data/code into a system or application to manipulate its operations. Eavesdropping Intercepting communications between two parties without their consent. Service Disruption Overloading or disrupting services to make them unavailable to legitimate users. Spoofing Impersonating a legitimate entity to deceive users or systems. Protocol-Based Attacks Exploiting inherent weaknesses or misconfigurations in com- munication protocols (e.g., TCP Slowloris) to overwhelm, dis- rupt services, or gain unauthorized access. Table 1 : Network attacks categories from the literature. These attack types form the basis for our testing framework, allowing us to simulate various attack scenarios and validate the robustness of network protocols under different adversarial conditions. (b) Attack Vectors. An attack vector is the specific method or pathway that an attacker uses to exploit a vulnerability in a system. Understanding attack vectors is essential in constructing effective security models and testing methodologies, as they directly influence how attacks are simulated and mitigated. We identified three main attack vectors relevant to network security testing, presented in Table 2 . Although malicious clients and servers might appear similar, the directions of their interactions differ. A malicious client begins communication and strays from protocols to perform attacks. In contrast, a malicious server responds to requests with harmful actions, taking advantage of client trust. The impact of a malicious client