Automated Backend Security Testing with AFL++ Fuzzer

In the last decades, fuzzing has shown to be an effective software vulnerabilities discovering technique. By subjecting a program to a multitude of diverse inputs, fuzzing primarily serves as a means of security testing. Despite efforts by the Open Source community to enhance usability, fuzzers mainly remain tools in the hands of security experts. Improvements are still needed to make them accessible to a wider audience. In Amadeus IT, security teams currently employ a manual process to deploy a black-box fuzzer over the network for testing critical systems. However, this approach proves to be both inefficient in scaling and ineffective in yielding high-quality results. In this thesis, we explore the current fuzzing research landscape to introduce a refined and more effective fuzzing solution. The focus is on coverage-guided fuzzers, which have the capability to dynamically craft new inputs based on feedback given by the execution of the target program. Additionally, we conduct an in-depth analysis of the company's technological stack to select a fuzzing solution that best aligns with the needs and limitations of the current environment. An engineering effort is undertaken to integrate AFLplusplus, the selected fuzzer, within the company's core C++ development environment. Furthermore, a Proof of Concept framework is implemented to allow developers to fuzz their applications by easily writing fuzz targets, without the need to involve the security teams. This solution not only improves the scalability of the fuzzer's usage across the company's applications but also enhances the effectiveness of the fuzzing process. The accomplishments of this internship set the stage for future company research, aiming to integrate fuzz testing directly within the application development lifecycle and surpass the outdated method.