The integration of blockchain into IoT can provide reliable control of the IoT network's
ability to distribute computation over a large number of devices. It also allows the AI
system to use trusted data for analysis and forecasts while utilising the available IoT
hardware to coordinate the execution of tasks in parallel, using a fully distributed
approach.
This thesis's rst contribution is a practical implementation of a real world IoT-
blockchain application,
ood detection use case, is demonstrated using Ethereum proof
of authority (PoA). This includes performance measurements of the transaction con-
rmation time, the system end-to-end latency, and the average power consumption.
The study showed that blockchain can be integrated into IoT applications, and that
Ethereum PoA can be used within IoT for permissioned implementation. This can be
achieved while the average energy consumption of running the
ood detection system
including the Ethereum Geth client is small (around 0.3J).
The second contribution is a novel IoT-centric consensus protocol called honesty-
based distributed proof of authority (HDPoA) via scalable work. HDPoA was analysed
and then deployed and tested. Performance measurements and evaluation along with
the security analyses of HDPoA were conducted using a total of 30 di erent IoT de-
vices comprising Raspberry Pis, ESP32, and ESP8266 devices. These measurements
included energy consumption, the devices' hash power, and the transaction con rma-
tion time. The measured values of hash per joule (h/J) for mining were 13.8Kh/J,
54Kh/J, and 22.4Kh/J when using the Raspberry Pi, the ESP32 devices, and the
ESP8266 devices, respectively, this achieved while there is limited impact on each de-
vice's power. In HDPoA the transaction con rmation time was reduced to only one
block compared to up to six blocks in bitcoin.
The third contribution is a novel, secure, distributed and decentralised architecture
for supporting the implementation of distributed arti cial intelligence (DAI) using
hardware platforms provided by IoT. A trained DAI system was implemented over the
IoT, where each IoT device hosts one or more neurons within the DAI layers. This
is accomplished through the utilisation of blockchain technology that allows trusted
interaction and information exchange between distributed neurons. Three di erent
datasets were tested and the system achieved a similar accuracy as when testing on a
standalone system; both achieved accuracies of 92%-98%. The system accomplished
that while ensuring an overall latency of as low as two minutes. This showed the secure architecture capabilities of facilitating the implementation of DAI within IoT
while ensuring the accuracy of the system is preserved.
The fourth contribution is a novel and secure architecture that integrates the ad-
vantages o ered by edge computing, arti cial intelligence (AI), IoT end-devices, and
blockchain. This new architecture has the ability to monitor the environment, collect
data, analyse it, process it using an AI-expert engine, provide predictions and action-
able outcomes, and nally share it on a public blockchain platform. The pandemic
caused by the wide and rapid spread of the novel coronavirus COVID-19 was used as
a use-case implementation to test and evaluate the proposed system. While providing
the AI-engine trusted data, the system achieved an accuracy of 95%,. This is achieved
while the AI-engine only requires a 7% increase in power consumption. This demon-
strate the system's ability to protect the data and support the AI system, and improves
the IoT overall security with limited impact on the IoT devices.
The fth and nal contribution is enhancing the security of the HDPoA through
the integration of a hardware secure module (HSM) and a hardware wallet (HW). A
performance evaluation regarding the energy consumption of nodes that are equipped
with HSM and HW and a security analysis were conducted. In addition to enhancing
the nodes' security, the HSM can be used to sign more than 120 bytes/joule and
encrypt up to 100 bytes/joule, while the HW can be used to sign up to 90 bytes/joule
and encrypt up to 80 bytes/joule. The result and analyses demonstrated that the HSM
and HW enhance the security of HDPoA, and also can be utilised within IoT-blockchain
applications while providing much needed security in terms of con dentiality, trust in
devices, and attack deterrence.
The above contributions showed that blockchain can be integrated into IoT systems.
It showed that blockchain can successfully support the integration of other technolo-
gies such as AI, IoT end devices, and edge computing into one system thus allowing
organisations and users to bene t greatly from a resilient, distributed, decentralised,
self-managed, robust, and secure systems.
