Lightweight Cryptographic Authentication Using Facial Landmarks and LFSR for IoT Security

This paper presents a novel lightweight cryptographic scheme designed to enhance IoT security by utilizing facial landmark values for efficient key generation. The system extracts facial landmarks using MobileNetV2 and leverages these values as seeds in a Linear Feedback Shift Register (LFSR) to generate pseudorandom keys. The Simon and Speck cryptographic algorithms are employed for the encryption and decryption processes. The system was implemented and tested on a Raspberry Pi to demonstrate its real-world applicability, with statistical tests such as the Chi-Square and Runs Up-Down tests verifying the randomness of the generated keys. The Chi-Square test achieved p-values above 0.05, confirming a uniform distribution, while the Runs Up-Down test demonstrated Z-values within the −1.96 to 1.96 range, supporting randomness at a 95% confidence level. The results indicate strong randomness and efficient key generation, producing 3471 unique keys from 20 facial landmark values. The lightweight cryptographic approach balances robust security with low computational overhead, making it ideal for resource-constrained IoT devices. This research contributes to ongoing efforts to secure IoT networks through optimized cryptographic techniques tailored for embedded systems and other limited-resource environments.