Secure Satellite Downlink Communication With Attention-Enhanced Transmitter Fingerprinting

As the internal nature of satellite networks such as the wide coverage, the low received power at the ground, and the radio broadcasting, detecting replay or spoofing attacks is an intractable challenge. Due to the limited onboard processing ability or leaked keys, a lot of satellites cannot secure downlink communication with cryptography. As a result, non-cryptographic methods based on fingerprint modeling have become popular. Nevertheless, existing methods overlook the difference between fingerprints during message transmission, leading to insufficient performance. To this end, we propose an attention-enhanced transmitter fingerprinting method (Att-Sat) to authenticate transmissions from satellites in a non-cryptographic way. We use the attention mechanism to make the model learn to pay more attention to the distinct fingerprint features of a specific transmitter, exploring more distinguished fingerprint characteristics for each transmitter. Experimental results on a publicly available dataset collected from the Iridium constellation indicate the effectiveness of our method. Under different settings, our method achieves a 6.2% improvement of AUC and a 6.5% improvement of EER on average compared with the state-of-the-art method. Moreover, our method Att-Sat achieves significant improvement under the real replay attack scenario.