Impact Conflict Detection of IoT Services in Multi-resident Smart Homes

TL;DR

The paper tackles detecting impact conflicts among IoT services in multi-resident smart homes by quantifying environmental impact via an STL-based signal-deviation framework and estimating resident preferences from historical, overlapping service events using DBSCAN. Conflict likelihood is computed by combining an impact score with preferential and temporal proximity, yielding a probabilistic measure $CL$. Experiments on the CASAS dataset show that incorporating preferences improves accuracy to about $90\%$, with temperature and illumination conflicts well-captured, while sound-related results are hindered by data gaps. The work contributes a novel triad of impact quantification, preference mining, and likelihood-based conflict detection, offering a practical pathway to resolve conflicts in shared smart-home environments, while acknowledging privacy, security, and scalability considerations for future work.

Abstract

We propose a novel impact conflict detection framework for IoT services in multi-resident smart homes. The proposed impact assessment model is developed based on the integral of a signal deviation strategy. We mine the residents' previous service usage records to design a robust preference estimation model. We design an impact conflict detection approach using temporal proximity and preferential proximity techniques. Experimental results on real-world datasets demonstrate the effectiveness of the proposed approach.

Figures (6)

• Figure 1: Service impact conflict: (a) progressive change, (b) instantaneous change.
• Figure 2: System architecture and conflict detection framework.
• Figure 3: Impact assessment: (a & d) robustness measurement, (b & e) percentage of impact time, (c & f) integral of deviation.
• Figure 4: Temperature distribution to get optimal preference range.
• Figure 5: Experimental results based on accuracy, precision, recall, f1-score.