The paper presents angr, a binary analysis framework that integrates state-of-the-art techniques for analyzing binary code. Binary analysis is crucial for identifying vulnerabilities in code that cannot be analyzed at the source level. However, many binary analysis techniques are research prototypes, leading to wasted effort and difficulty in comparing approaches. angr addresses this by providing a unified framework that allows researchers to compose and compare techniques. The framework is open-sourced and includes building blocks for various analyses, enabling the development of new approaches. The paper discusses the challenges of automated binary analysis, including the trade-offs between replayability and semantic insight. It also highlights the DARPA Cyber Grand Challenge, a competition that provided a dataset of binaries for evaluating analysis techniques. This dataset allows for the comparison of different approaches in vulnerability discovery, exploit generation, and software patching. The paper describes the implementation of angr, a next-generation binary analysis engine, and presents several offensive analyses that demonstrate its effectiveness. These analyses include vulnerability discovery, exploit replaying, automatic exploit generation, and exploit hardening. The framework enables the combination of different techniques to enhance their capabilities. The paper also discusses the challenges of dynamic symbolic execution, including the path explosion problem, and proposes solutions such as symbolic-assisted fuzzing and under-constrained symbolic execution. These techniques aim to balance the trade-offs between scalability and semantic insight. The paper concludes that angr provides a flexible and powerful framework for binary analysis, enabling the reproduction and improvement of existing techniques while supporting the development of new ones. The framework is designed to be user-friendly, with a Python-based implementation that allows for easy integration and use by the security community.The paper presents angr, a binary analysis framework that integrates state-of-the-art techniques for analyzing binary code. Binary analysis is crucial for identifying vulnerabilities in code that cannot be analyzed at the source level. However, many binary analysis techniques are research prototypes, leading to wasted effort and difficulty in comparing approaches. angr addresses this by providing a unified framework that allows researchers to compose and compare techniques. The framework is open-sourced and includes building blocks for various analyses, enabling the development of new approaches. The paper discusses the challenges of automated binary analysis, including the trade-offs between replayability and semantic insight. It also highlights the DARPA Cyber Grand Challenge, a competition that provided a dataset of binaries for evaluating analysis techniques. This dataset allows for the comparison of different approaches in vulnerability discovery, exploit generation, and software patching. The paper describes the implementation of angr, a next-generation binary analysis engine, and presents several offensive analyses that demonstrate its effectiveness. These analyses include vulnerability discovery, exploit replaying, automatic exploit generation, and exploit hardening. The framework enables the combination of different techniques to enhance their capabilities. The paper also discusses the challenges of dynamic symbolic execution, including the path explosion problem, and proposes solutions such as symbolic-assisted fuzzing and under-constrained symbolic execution. These techniques aim to balance the trade-offs between scalability and semantic insight. The paper concludes that angr provides a flexible and powerful framework for binary analysis, enabling the reproduction and improvement of existing techniques while supporting the development of new ones. The framework is designed to be user-friendly, with a Python-based implementation that allows for easy integration and use by the security community.