SSARS: Secure smart-home activity recognition system

Abstract

Smart homes provide assistance services that enhance the well-being, independence, and health of the residents, particularly the elderly. As techniques for human activity recognition in smart homes continue to advance, current methods face challenges such as insecure transmission of raw data and individual movement classification. To overcome these challenges, this study proposes Secure Smart-Home Activity Recognition System (SSARS). The proposed methodology utilizes an advanced preprocessing technique, AI-PSD, to reduce impulse noise in the data by combining adaptive interpolation (AI) and power spectral density (PSD). The Fractional Fast Fourier Transform (F-FFT) effectively captures statistical and dynamic aspects of human activities, offering a more detailed understanding of movement patterns. The extracted features are securely transmitted through encryption based on Factor private Key-based Elliptic Curve Cryptography (FK-ECC). Additionally, this study introduces the Pade activation function with a modified Physical Neural Network (P-PNN) to improve the system's classification ability. The proposed SSARS showed outstanding performance across various metrics, including an accuracy of 98.68 % and a precision of 98.93 % when compared with existing state-of-the-art approaches.