Abstract:
Fog computing defines an architecture-driven, decentralised model dispersing processing, applications,
and storage between cloud infrastructure and IoT data sources. While scalability
and latency are reduced, security and load balancing remain issues. This research addresses
these challenges by proposing a load-balancing framework, implemented in an environment of
FogBus, involving blockchain to guarantee authenticity and encryption of sensitive data. The
framework operates across three layers: IoT to Fog, Blockchain Implementation, and Fog to
Cloud, enabling communication security and optimal algorithm running across devices. There
are several blockchain integration methods, including having IoT devices as whole blockchain
nodes to support direct communication, utilising gateways to filter, compress, and encrypt data
before it’s transmitted, and utilising sidechains to reduce load on the root blockchain. Transactions
and communications happen automatically through smart contracts, and efficiency and
reduction in manual intervention are increased. IoT devices have robust security capabilities to
protect against potential threats, and the integrity of data and trust throughout the network is
upheld. The framework put forward here was tested and proven in a Dockized testbed in two
modes: fully cloud and hybrid cloud–fog. Simulated IoT devices generated workloads to monitor
latency, execution time, network bandwidth, and message throughput. We observe that the
hybrid cloud–fog architecture outperformed, in all cases, the all-cloud architecture, and average
message throughput varied from 0.9937 to 0.9993 messages/second. The IoT to Fog communication
path had minimum delay (2–19 ms) and linear delay growth upon increase in sensors,
and IoT–Cloud through QoS had maximum delays (6–70 ms) and steepest drop in performance.
These results confirm that fog nodes and blockchain security boost trust, reduce latency, and
raise system scalability. The system is most suited to latency-sensitive, security-aware applications
such as medical diagnostics, military alarm systems, and weather monitoring, and future
work will focus on in-the-wild verification and integration of machine learning to support predictive
resource provisioning.
