https://doi.or g/10.31449/inf.v48i1.5739 Informatica 48 (2024) 141–142 141
Enabling Decentralized Privacy Pr eserving Data Pr ocessing
in Sensor Networks
Niki Hrovatin
Faculty of Mathematics, Natural Sciences and Information T echnologies University of Primorska
E-mail: niki.hrovatin@famnit.upr .si
Thesis summary
Keywords: sensor networks, privacy , onion routing, distributed computing, multy-party computation
Received: February 15, 2024
The paper summarizes the findings of the Doctoral Thesis [ 1 ]. W e pr opose a paradigm shift fr om traditional
privacy-pr eserving joint computation, which r elies on data obfuscation methods, to privacy pr eservation
thr ough anonymity . The main contribution of the thesis is a privacy-pr eserving pr otocol based on the Onion
Routing concept that allows sensor network nodes to jointly compute an arbitrary function and keeps the
participating nodes and their inputs private. W e demonstrate the pr otocol’ s security and, thr ough simula-
tions, its effectiveness in lar ge sensor networks.
Povzetek: Doktorska disertacija pr edlaga novo metodo ohranjanja zasebnosti pr eko anonimnosti, s
poudarkom na pr otokolu za ohranjanje zasebnosti, osnovanem na konceptu Onion Routinga, ki omogoča
skupno izračunavanje funkcij v omr ežjih senzorjev , pri čemer ohranja zasebnost sodelujočih vozlišč in nji-
hovih vhodov .
1 Intr oduction
In today’ s technological landscape, Sensor Networks are
crucial for capturing geographically spread physical phe-
nomena,serving a broad spectrum of applications from en-
vironmental monitoring to industrial automation. Despite
their benefits, sensor networks also have several limitations
such as susceptibility to faults, limited processing capacity ,
and vulnerabilities to security and privacy breaches [ 2 ].
These limitations are particularly prominent in the tradi-
tional centralized sensor network architecture, where nodes
collect and transmit raw data to a remote system outside
the sensor network for processing and analysis. As a result,
there is a shift towards decentralized architectures, driven
by the edge computing paradigm, performing data process-
ing in the sensor network as close as possible to the data
source [ 3 ]. Despite the benefits of edge computing and de-
centralization, existing distributed computing frameworks
for sensor networks lack universality and face issues with
security , privacy and ef ficiency . Specialized for tasks like
data aggregation, query processing or machine learning,
these frameworks struggle with adaptability .
This paper presents a summary of a Doctoral Thesis [ 1 ],
introducing a novel communication protocol [ 4 ] that en-
ables the joint computation of arbitrary functions on sensor
network nodes and keeps the participating nodes and their
inputs private.
2 The communication pr otocol
The communication protocol is based on the Onion Routing
technique for anonymous communication over a computer
network. W e similarly employ messages structured into en-
cryption layers, such that a layer can be decrypted only by
the tar geted node revealing an inner encryption layer ad-
dressed to another node in the network. Therefore, message
decryption is carried out gradually by leading the layered
message across network nodes following the precise order
given at message construction.
Encryption layers are not enclosing only the inner layer ,
but also additional secret information revealed only to the
node decrypting that layer . Path details and encryption keys
are in this way conveyed to in-path nodes. Specifically , en-
cryption key pairs, are delivered only to a subset of nodes in
the message path. Unlike traditional onion routing, where
encryption keys establish an anonymous communication
channel, here, the keys grant access to the payload contain-
ing edge computing information. Please note that pairs of
symmetric encryption keys include distinct keys; however ,
pairs are chained through layers of the layered object, as
can be seen from Fig. 1 .
The described protocol ensures privacy by establishing
an anonymity set that conceals the nodes accessing the pay-
load among all the nodes in the message path.

142 Informatica 48 (2024) 141–142 N. Hrovatin
sink node
sensor node
IP:123
IP:237
IP:42
IP:11
IP:22
IP:877
s
a
s
b
next hop ip pair of symmetric encryption keys
public key encryption
main()
symmetric encryption
binary string task
101...1 Head:
QUERY
Payload:
s
b
s
c IP:237
IP:42
IP:877
IP:22
ID
query identifier
Figure 1: Illustration of messages defined by the privacy-preserving communication protocol.
3 Evaluation methodology and
r esults
W e provided privacy preservation analysis and formal
proofs showing that the protocol is secure against the ex-
ternal and internal attacker models.
W e realized a simulation of the protocol using the ns-
3 simulator
1
, testing it with networks of up to 400 nodes
across two network topologies and testing several protocol
parameters. Results show that the protocol is scalable and
adequate for application in sensor networks.
The protocol was tested for machine learning training
and inference. Results show that models trained using
the protocol achieve comparable performance to machine
learning models trained using traditional batch learning.
4 Discussion and further work
Our results demonstrate the protocol’ s ef fectiveness in pre-
serving privacy , its high adaptability to various data pro-
cessing tasks and the feasibility of application in lar ge-scale
sensor networks. Moving forward, we plan to transition our
protocol from theory to practice by implementing it in real-
world settings to collect and analyze air quality data directly
on-site. Additionally , we plan to extend our protocol’ s ap-
plication to the broader Internet of Things, in the form of
a permission-less decentralized resource marketplace that
incentivizes user participation and leverages blockchain for
trust.
Refer ences
[1] N. Hrovatin, Omogočanje decentralizirane obdelave
podatkov z var ovanjem zasebnosti v senzorskih om-
r ežjih: doktorska disertacija . PhD thesis, Univerza
1
Network simulator ns-3: https://www.nsnam.org/
na Primorskem, Fakulteta za matematiko, naravoslovje
in …, 2023.
[2] I. T omić and J. A. McCann, “A survey of potential se-
curity issues in existing wireless sensor network proto-
cols,” IEEE Internet of Things Journal , vol. 4, no. 6,
pp. 1910–1923, 2017. https://doi.org/10.1109/
JIOT.2017.2749883 .
[3] A. Sorniotti, L. Gomez, K. W rona, and L. Odorico, “Se-
cure and trusted in-network data processing in wireless
sensor networks: a survey ,” Journal of Information As-
surance and Security , vol. 2, no. 3, pp. 189–199, 2007.
[4] N. Hrovatin, A. T ošić, M. Mrissa, and J. V ičič, “A gen-
eral purpose data and query privacy preserving proto-
col for wireless sensor networks,” IEEE T ransactions
on Information For ensics and Security , 2023. https:
//doi.org/10.1109/tifs.2023.3300524 .