Content uploaded by Jorge Peña Queralta
Author content
All content in this area was uploaded by Jorge Peña Queralta on Apr 20, 2020
Content may be subject to copyright.
Edge AI and Blockchain for
Privacy-Critical and Data-Sensitive Applications
A. Nawaz1,2, T. N. Gia2, J. Pe˜
na Queralta2and T. Westerlund2
1Shanghai Key Laboratory of Intelligent Information Processing, Fudan University, China
2Turku Intelligent and Embedded Robotic Systems (TIERS), University of Turku, Finland
Emails: 1{nanum18, hbkan}@fudan.edu.cn, 2{jopequ, tunggi, tovewe}@utu.fi
Abstract—The edge and fog computing paradigms enable
more responsive and smarter systems without relying on cloud
servers for data processing and storage. This reduces network
load as well as latency. Nonetheless, the addition of new layers
in the network architecture increases the number of security
vulnerabilities. In privacy-critical systems, the appearance of
new vulnerabilities is more significant. To cope with this issue,
we propose and implement an Ethereum Blockchain based
architecture with edge artificial intelligence to analyze data at the
edge of the network and keep track of the parties that access the
results of the analysis, which are stored in distributed databases.
Index Terms—Blockchain; Edge Computing; AI; E-Health;
U-Health; IoT; Internet of Things; Fall Detection; Ubiquitous
Health; Ethereum;
I. INTRODUCTION
Users and organizations are becoming increasingly aware
of the importance and significance of protecting personal data
and online privacy. This is a particularly critical issue in the
IoT, where numerous security challenges have been identified
by the research community. In recent years, a wide variety
of IoT platforms and applications have adopted the use of
blockchain technology to mitigate multiple privacy risks and
allow secure transactions without the need for a trusted party.
Nevertheless, current integrations of blockchain within the
IoT have been focusing on securing communication without
changing the interaction topology. Exploiting the fog and
edge computing paradigms, we propose an extension of the
Ethereum blockchain to resource-constrained devices. With
our proposed platform, end-devices can negotiate directly with
third parties regarding the use of their data. This ensures
data owners are always aware of transactions involving their
data. In addition, because of the immutable nature of the
blockchain, all transactions are recorded and auditable, which
further reduces the possibilities of misuse of private data. We
have implemented and validated the proposed platform in a
real application, demonstrating its potential for integration of
IoT devices with scarce computational capabilities. To enhance
privacy-critical systems, edge based AI techniques has been
implemented to restrict raw data to its producers only. But,
this domain still lacks the owner control over their sensitive
health data, where owner can process sensitive information
by using neural networks and sell statistics to the interested
clients. Furthermore, this reduces the network load and the
latency of Blockchain transactions [1], [2].
II. RE LATE D WORK
To make data access policies accessible at each level re-
searchers proposed blockchain based systems integrated with
edge computing. By implementing AI at edge nodes further
decreased privacy vulnerabilities. Mamoshina. P et. al pro-
posed access policies to accelerate the private patients data
and implement deep learning algorithms to turns raw data
into strong useful information which can be used in bigger
perspectives [1]. In [3], Mackey et. al proposed blockchain
based data privacy control opportunities and challenges which
are significant enough in healthcare applications. A similar
approach was presented by Peterson et. al [4].
III. ETHEREUM BLOCKCHAIN WITH EDGE AI
To exclude intermediaries involve in data transactions in
edge devices, we define a platform in which the Blockchain
paradigm is extended into scarce computing devices. Ethereum
blockchain is used as a service platform to run smart contracts
to make the system autonomous in therms of its’ commu-
nication, processing and data dealing. A private ethereum
blockchain network is created by creating a genesis file. To
add every device, a pair of private and public key is generated
which will be used as a identifier of a device.
In our proposed system, all resource constrained sensor
nodes are directly connected to and rely on the edge gate-
ways which are often implemented by powerful single board
computers able to work as miner nodes to store, analyze and
aggregate raw data. Miner nodes can run neural networks
to process the raw data received from sensor nodes. With
a predefined time interval, edge nodes process the raw data,
and save this processed information into a new data block
by creating a unique hash. This data block consists of two
parts, header and body. The body apart contains processed
information and header part consists of general characteristics
about the processed information. This includes hash of previ-
ous block, time stamp, raw data definition and the type of data,
which can be further use for combining heterogeneous data at
bigger level for the sake of intelligent systems. To protect
the hash of data block, symmetric cryptography is used. After
encryption, the data blocks are saved on a blockchain cloud
and key is only hold by the end-device. Which will be later
used by a client to decrypt the desired data. Moreover, every
access to the data will be recorded. The proposed architecture
is illustrated in Figure 1, which is composed of four layers.
SENSORLAYER EDGELAYER-BLOCKCHAINNODES CLOUDLAYER END-USER
APPLICATION
Global Storage for Encrypted Data
Cloud Services
Web/Mobile Application Servers
Bio-signals analysis
ECG Feature Extraction and Storage in the Blockchain
Fig. 1. Proposed System Architecture
0 0.511.5 2
(a) Fragment of raw data over 2.3 seconds
(b) Extracted cycle template
2 4 6 8 10
60
80
100
(c) Heart rate over 10 seconds
Fig. 2. Results of the data analysis at the edge gateway.
ECG TCT RT TRD
150
72
36 35
Fig. 3. Execution time of the different processes (ms).
IV. EXP ER IM EN T AN D RES ULTS
The raw ECG data collected from a healthy 30 year-old
male person is shown in Figure 2. The data is sent to a smart
Edge-assisted gateway which extracts different ECG features
such as heart rate [5]. We have utilized a Raspberry Pi model
3 as the edge gateway, which in turn runs a node of the
Ethereum Blockchain. In order to test the feasibility of the
proposed model, we accumulate data for 10 seconds and then
analyze it. The data analysis requires around 150ms for the
feature extraction. Then, the results are encrypted and stored
in a distributed storage solution. The metadata is stored in
the blockchain. Figure 3 shows the execution time of the
analysis process (ECG), a data retrieval transaction (TRD),
a transaction confirmation (TCT) and the response time (RT).
In total, the system needs around 300ms to process one batch
of data, which runs every 10 seconds. Therefore, one gateway
could support up to 20 or 30 end-devices with the proposed
architecture.
V. CONCLUSION AND FUTURE WORK
Integrating Blockchain with Edge computing opens new
paradigms in privacy-critical and data-sensitive applications.
Our proposed architecture, combining a distributed ledger
with AI at the edge, creates secure database of processed
information which can only be used with the permission of its
owner. By Edge AI we refer to local decision making and data
processing at the edge computing layer. End devices can di-
rectly control all the processing, analyzing and sharing of their
data by updating their policies via ethereum smart contracts.
Implementing AI at edge nodes reduces resource consumption
like bandwidth required to upload data to blockchain cloud as
well as local storage. This data analysis step also increases
privacy by storing only processed information rather than raw
data.
In future work, we will analyze the scalability of the
proposed approach and alternative applications and experiment
with the integration of more complex deep learning algorithms.
We will study the utilization of Ethereum 2 for more scalable
systems.
REFERENCES
[1] P. Mamoshina et al. Converging blockchain and next-generation artifi-
cial intelligence technologies to decentralize and accelerate biomedical
research and healthcare. Oncotarget, 9(5):5665, 2018.
[2] J. Pe˜
na Queralta et al. Edge-AI in LoRa-based healthcare monitoring:
A case study on fall detection system with LSTM Recurrent Neural
Networks. In 42nd TSP, 2019.
[3] T. K. Mackey et al. ‘fit-for-purpose?’–challenges and opportunities for
applications of blockchain technology in the future of healthcare. BMC
medicine, 17(1):68, 2019.
[4] K. Peterson et al. A blockchain-based approach to health information
exchange networks. In NIST W. Blockchain Healthcare, 2016.
[5] C. Carreiras et al. BioSPPy: Biosignal processing in Python, 2015–.
[Online; accessed Aug. 2019].
