Socially Grounded IoT Protocol for Reliable Computer Vision in Industrial Applications

Chidambaram, Gokulnath and Subbarayappa, Shreyanka and Sai Baba, M (2026) Socially Grounded IoT Protocol for Reliable Computer Vision in Industrial Applications. Future Internet, 18 (2).

Text
Socially Grounded IoT Protocol for Reliable Computer Vision in Industrial Applications.pdf - Published Version
Download (5MB)

Contribution	Name	Email

Abstract:	The Social Internet of Things (SIoT) enables collaborative service provisioning among interconnected devices by leveraging socially inspired trust relationships. This paper proposes a socially driven SIoT protocol for trust-aware service selection, enabling dynamic friendship formation and ranking among distributed service-providing devices based on observed execution behavior. The protocol integrates detection accuracy, round-trip time (RTT), processing time, and device characteristics within a graph-based friendship model and employs PageRank-based scoring to guide service selection. Industrial computer vision workloads are used as a representative testbed to evaluate the proposed SIoT trust-evaluation framework under realistic execution and network constraints. In homogeneous environments with comparable service-provider capabilities, friendship scores consistently favor higher-accuracy detection pipelines, with F1-scores in the range of approximately 0.25–0.28, while latency and processing-time variations remain limited. In heterogeneous environments comprising resource-diverse devices, trust differentiation reflects the combined influence of algorithm accuracy and execution feasibility, resulting in clear service-provider ranking under high-resolution and high-frame-rate workloads. Experimental results further show that reducing available network bandwidth from 100 Mbps to 10 Mbps increases round-trip communication latency by approximately one order of magnitude, while detection accuracy remains largely invariant. The evaluation is conducted on a physical SIoT testbed with three interconnected devices, forming an 11-node, 22-edge logical trust graph, and on synthetic trust graphs with up to 50 service-providing nodes. Across all settings, service-selection decisions remain stable, and PageRank-based friendship scoring is completed in approximately 20 ms, incurring negligible overhead relative to inference and communication latency.
Item Type:	Journal Paper
Subjects:	School of Natural and Engineering Sciences > Others
Divisions:	Schools > Natural Sciences and Engineering
Date Deposited:	03 Mar 2026 08:53
Last Modified:	03 Mar 2026 10:36
Official URL:	https://www.mdpi.com/1999-5903/18/2/69
Related URLs:
Funders:	*
Projects:	*
DOI:	https://doi.org/10.3390/fi18020069
URI:	http://eprints.nias.res.in/id/eprint/3261

Actions (login required)

View Item