gr-PHYSEC: Real-time Channel-based Key Generation for Physical Layer Secure Wireless Communications

A GNU Radio OOT module leveraging neural network channel feature extraction, Reed-Solomon reconciliation, and SHA-512 hashing for physical-layer key generation

Jose Angel Sanchez Viloria; George Sklivanitis; Dimitris Pados· 2026· DOI 10.48550/arXiv.2609.16375

The core problem

Securing wireless communication against eavesdropping is critical, particularly in dynamic and decentralized environments. Traditional key generation relies on pre-shared secrets or computational complexity, which may be impractical or vulnerable in such settings. gr-PHYSEC introduces a new GNU Radio out-of-tree (OOT) module for real-time physical-layer key generation. Unlike traditional approaches, it derives symmetric keys from the wireless channel's inherent randomness. The system embeds a trained neural network within GNU Radio to extract channel features between trusted parties (Alice and Bob) during probe exchanges. These features are quantized into binary keys, reconciled via Reed-Solomon encoding, and further secured with SHA-512 hashing. The generated keys are then directly used to encrypt data. This work showcases how GNU Radio can support real-time AI-driven security solutions, pushing the boundaries of software-defined secure communication.

Innovation

Real-world experiments were conducted at the FAU CAAI connected robotics testbed using ADALM Pluto software-defined radios and NVIDIA Jetson Orin platforms mounted on ground robotic platforms. The performance of gr-PHYSEC was evaluated in terms of key disagreement rate (KDR) and randomness. - **Key disagreement rate**: The system achieved low KDR, indicating that Alice and Bob generated identical keys with high probability after reconciliation. The exact KDR values are not provided in the abstract, but the results demonstrate low disagreement. - **Randomness**: The generated keys passed the NIST test suite for random and pseudorandom number generators for cryptographic applications, confirming strong randomness. - **Real-time operation**: The integration with GNU Radio and the Jetson Orin enabled real-time key generation and encryption during robotic communication. The experiments validate the feasibility of physical-layer key generation in dynamic environments with mobile robots. The use of neural networks for feature extraction proved effective in capturing channel randomness while mitigating noise and interference.

Securing wireless communication against eavesdropping is critical, particularly in dynamic and decentralized environments. Traditional key generation relies on pre-shared secrets or computational complexity, which may be impractical or vulnerable in such settings. gr-PHYSEC introduces a new GNU Radio out-of-tree (OOT) module for real-time physical-layer key generation. Unlike traditional approaches, it derives symmetric keys from the wireless channel's inherent randomness. The system embeds a trained neural network within GNU Radio to extract channel features between trusted parties (Alice and Bob) during probe exchanges. These features are quantized into binary keys, reconciled via Reed-Solomon encoding, and further secured with SHA-512 hashing. The generated keys are then directly used to encrypt data. This work showcases how GNU Radio can support real-time AI-driven security solutions, pushing the boundaries of software-defined secure communication.

The gr-PHYSEC module operates in real time within the GNU Radio framework. The key generation process consists of several stages:

Why it matters

The gr-PHYSEC module demonstrates a practical approach to physical-layer security by leveraging the inherent randomness of wireless channels. Unlike computational key exchange, which relies on the hardness of mathematical problems, this method exploits the physical characteristics of the channel, making it information-theoretically secure against eavesdroppers with limited channel knowledge. The integration of a neural network within GNU Radio enables real-time operation and adaptability to changing channel conditions. The use of Reed-Solomon reconciliation and SHA-512 hashing ensures that the final keys are identical between Alice and Bob and possess strong randomness. The NIST test suite validation provides confidence in the cryptographic quality of the keys. The experiments on ground robots highlight the applicability of this approach in decentralized, dynamic environments such as robot swarms or tactical networks. However, the system's performance depends on the channel's randomness and the eavesdropper's location. If Eve is close to Alice or Bob, the channel correlation may be high, reducing security. Future work could explore adaptive quantization and advanced neural network architectures to improve robustness. Additionally, the computational overhead of the neural network and hashing on resource-constrained devices like the Jetson Orin should be further optimized. Overall, gr-PHYSEC represents a significant step toward AI-driven physical-layer security in software-defined radio platforms. The source code is available at https://github.com/C2A2-at-Florida-Atlantic-University/gr-PHYSEC, facilitating further research and development.

Who should read this

CS practitioners and researchers

Opening member content…