Modern wireless communications systems are constantly evolving and growing more complex. Recently, there has been a shift towards software defined radios due to the flexibility soft- ware implementations provide. This enables an easier development process, longer product lifetimes, and better adaptability for congested environments than conventional hardware systems. However, this shift introduces new attack surfaces where vulnerable implementa- tions can be exploited to disrupt communications or gain unauthorized access to a system. Previous research concerning wireless security mainly focuses on vulnerabilities within pro- tocols rather than in the radios themselves. This dissertation specifically addresses this new threat against software radios and introduces a new security model intended to mitigate this threat. We also demonstrate example exploits of waveforms which can result in either a denial-of-service or a compromise of the system from a wireless attack vector. These example exploits target vulnerabilities such as overflows, unsanitized control inputs, and unexpected state changes.

We present a defense-in-depth security architecture for software radios that protects the system by isolating components within a waveform into different security zones. Exploits against vulnerabilities within blocks are contained by isolation zones which protects the rest of the system from compromise. This architecture is inspired by the concept of a microkernel and provides a minimal trusted computing base for developing secure radio systems. Unlike other previous security models, our model protects from exploits within the radio protocol stack itself and not just the higher layer application. Different isolation mechanisms such as containers or virtual machines can be used depending on the security risk imposed by a component and any security requirements. However, adding these isolation environments incurs a performance overhead for applications. We perform an analysis of multiple example waveforms to characterize the impact of isolation environments on the overall performance of an application and demonstrate the overhead generated from the added isolation can be minimal. Because of this, our defense-in-depth architecture should be applied to real-world, production systems. We finally present an example integration of the model within the GNU Radio framework that can be used to develop any waveform using the defense-in-depth se- curity architecture.