Towards developing an actor-based immune system for smart homes

Abstract

Smart home environments receive substantial improvement from the Internet of Things (IoT) through automated systems and connected devices, which optimize living space management. The advanced technology of smart homes requires strong anomaly detection systems together with Root Cause Analysis (RCA) to maintain security and reliability. This paper presents a new immune system model for smart homes that detects unusual behavior patterns and conducts full RCA. Our methodology uses the Actor Model together with deep learning approaches that process sensor events while applying causal inference to detect anomalies and their root causes. We use multiple deep learning architectures, including sequence-to-sequence (Seq2Seq), autoencoder, and LSTM networks, to detect various anomalies, which include missing data and abnormal data values. Our autoencoder-based solution demonstrates superior performance, achieving 96.2 % precision and a 98.0 % F1-score. These results represent a significant improvement of 76.8 % in precision and 73.7 % in F1-score over state-of-the-art baseline methods. Our research demonstrates how advanced techniques improve both anomaly detection accuracy and the efficiency of RCA, which results in better smart home environment reliability and resilience.