Quantum Anomaly Detection for Advanced Persistent Threats Using
Quantum Support Vector Machines
Jaideep Rukmangadan and Seema Vasudevan
Deparment of Mechatronics Engineering, The Oxford College of Engineering, Bangalore, Karnataka, India
Keywords: Cybersecurity, Anomaly Detection, Encrypted Communication, Advanced Persistent Threats (APTs),
Quantum Support Vector Machines (QSVM), Quantum Kernel Methods, Quantum Machine Learning.
Abstract: Cyber threat sophistication, in particular Advanced Persistent Threats (APTs), requires new detection
technologies to deal with big and encrypted data. Anomaly detection has become an unsustainable process in
big data environments with encryption making them even more so. This article presents a new method
utilizing QSVM (Quantum Support Vector Machines) and Quantum Kernel Methods to find out anomalies in
encrypted communication paths. Quantum kernels can translate input information into higher-dimensional
Hilbert spaces, with computational efficiency and precision over and above that of the classical methods. The
use of QSVM can detect faint signals from APTs (like Zero-Day attacks) more accurately. This paper also
tests the security of encryption protocols such as RSA and AES on quantum simulators and proposes quantum-
safe alternatives to protect against quantum attacks before they happen. Experimental findings show
significant enhancements in anomaly detection performance and computation speed, which are a first of their
kind for quantum-based cybersecurity systems.
1 INTRODUCTION
Advanced Persistent Threats (APTs) are the most
perilous and hard to predict cyberattack type, which
are products of extremely smart attackers who find
vulnerabilities for longer. Attacks are even harder to
identify since they go undetected within encrypted
communication systems. Models of anomaly
detection using the classic machine learning
algorithms are severely restricted. Encrypted data
requires a lot of computing power and typically
decryption which involves privacy risks and waste.
Quantum Computing promises a silver bullet to solve
these issues. Quantum algorithms, like Quantum
Support Vector Machines (QSVM) and Quantum
Kernel Methods developed and refined in 2024 offer
exponential computational benefits over the old
methods. But QSVM, unlike traditional SVMs, can
use quantum superposition to look at the data patterns
in multiple dimensions at the same time. It also gives
you an extra edge when it comes to anomaly
detection, where you can catch anomalies faster and
more accurately in encrypted traffic, which is where
the conventional methods just don’t do a good job.
This paper describes a QSVM-based APT detection
system in real time without compromising on data
privacy. Incorporating quantum kernels, the
technique encrypts communication and translates it
into higher-dimensional space so that anomalies are
caught in time. We perform anomaly detection as well
as testing whether existing encryption algorithms
such as RSA and AES are resilient to quantum attacks
with quantum simulators. This study points to both
the weaknesses of existing encryption standards and
the ability of quantum-resilient cryptography to
augment cybersecurity systems.
2 RELATED WORKS
R. Alluhaibi., 2024; J. D. Bakos., 2023; O. Faker and
N. E. Cagiltay, 2023. The rapid advancements in
cybersecurity, particularly in anomaly detection and
encrypted data analysis, have seen significant
research contributions in recent years. Existing works
on classical machine learning techniques such as
Support Vector Machines (SVM), Convolutional
Neural Networks (CNN), and Long Short-Term
Memory (LSTM) have demonstrated moderate
success in identifying cyber threats and Advanced
Persistent Threats (APTs) within networks. M. J. H.
Rukmangadan, J. and Vasudevan, S.
Quantum Anomaly Detection for Advanced Persistent Threats Using Quantum Support Vector Machines.
DOI: 10.5220/0013896400004919
Paper published under CC license (CC BY-NC-ND 4.0)
In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 3, pages
261-266
ISBN: 978-989-758-777-1
Proceedings Copyright © 2025 by SCITEPRESS – Science and Technology Publications, Lda.
261

Faruk, et al., 2022; C. Kenyon and C. Capano, 2022,
However, such models are computationally
expensive (for large scale encrypted data sets) and
rarely decrypt encrypted communications which
compromise privacy. And classical methods can’t
effectively detect fine and complex patterns in
encrypted data flows due to features representation
limitation. R. Kharsa., et al, 2023; D. Lakshmi., et al.,
2023 Quantum computers, oh-so-potent, recently
transformed anomaly detection paradigms. There has,
for example, been research on using quantum
machine learning models Quantum Support Vector
Machines (QSVM) for cybersecurity tasks. Quantum
models make use of quantum superposition and
kernel operations on high-dimensional data with
computationally better performances than traditional
approaches. M. Macas., et al, 2022; D. Said., 2023
But despite being promising, current quantum-based
work has been mainly theoretical in nature or has
been restricted to simplified data, thus not
generalisable. S. K. Sheoran and V. Yadav, Also,
scalability is a problem with current quantum
solutions as the majority of them are developed only
for small data sets without testing the robustness in
big, locked systems. S. K. Sood and M. Agrewal,
2024; K. Shara, 2023; W. S. Admass, et al, 2023 The
deficiencies in quantum attack resistance of
encryption were also explored recently in recent
publications revealing weakness of common
cryptographic standards like RSA and AES. M. S.
Akter, et al, 2023; Z. Ali, et al., 2022 Classical
encryption methods are secure enough against current
attack but they have been hampered by quantum
decryption techniques with increasing quantum
circuit depths and computing efficiencies. R.
Alluhaibi, 2024; J. D. Bakos, 2023 However, few
experiments have demonstrated strong solutions to
overcome these problems and there is a gaping hole
in quantum-resistant encryption. The paper bypasses
such limitations by introducing an overall Quantum
Support Vector Machine (QSVM) solution for
anomaly detection in encrypted flow
communications. In contrast to models currently in
use, QSVM can operate directly on encrypted data
without decryption – this is more privacy and
security. E. F. Combarro, 2023; O. Faker and N. E.
Cagiltay, 2023, By using quantum kernels, the
solution encodes the data into high-dimensional
feature spaces allowing the identification of small
anomalies and APTs very accurately. M. J. H. Faruk,
et al., 2022 Also, the model’s scalability is
extensively tested over large datasets, which solves
the problem of previous research which never
assessed quantum solutions under practical
conditions. As an add-on to anomaly detection, this
article proposes a quantum-resilient encryption
testing scheme that reveals the weaknesses of existing
cryptographic standards and also proposes quantum-
safe key exchange techniques to help reduce post-
quantum risks. Compared to classical and current
quantum models, the proposed framework is
computationally faster, more accurate and scalable –
and private in encrypted contexts. These advances fill
the most gaping holes in the existing literature, and
make the work a key piece of quantum-powered
cybersecurity research.
3 METHODOLOGY
It’s running Quantum Support Vector Machines
(QSVM) to identify anomalies in encrypted
communications real-time. QSVM works by
Quantum Kernel Methods, not traditional kernels like
a SVM to transform data into quantum feature space.
Quantum kernels are computed by quantum circuits
mapping input into high-dimensional space easily.
This transformation reinforces the normal vs.
abnormal pattern discrimination that’s key to sniffing
out non-obvious APTs. QSVM implementation starts
with encryption of data streams for use with quantum
feature encodings. The input is then used as a map
using quantum kernels into a Hilbert space where
anomalies are detected with optimised hyperplanes.
The quantum model utilises superposition, and so it is
able to analyse patterns simultaneously that classical
models analyze in one direction at a time. The very
parallelism of this system means much less
computation and resource cost.
Figure 1: An Intended Quantum-Resilient Encryption Test
Plan.
ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,
COMMUNICATION, AND COMPUTING TECHNOLOGIES
262

Figure 1 show the workflow for testing and
measuring the quantum immunity of newer
encryption protocols. This involves first simulations
of quantum cyberattacks on input encryption
algorithms (RSA/AES). The second resilience
assessment finds vulnerabilities in existing
cryptographic protocol, and returns vulnerability
scores. These discoveries go into the proposed
quantum-safe key exchange algorithms, resulting in
strong, quantum-secure encryption solutions. This
workflow provides an organized approach for
diagnosing and remediating encryption
vulnerabilities in advance of post-quantum cyber-
attacks.
Figure 2: QSVM-Based Anomaly Detection Framework.
Figure 2 Overview of QSVM anomaly detection
model for APT detection in encrypted
communication logs. The model starts with encrypted
data mapping into quantum feature space through
quantum kernel to represent higher-dimensional data.
The derived features are fed to the Quantum Support
Vector Machine (QSVM) for detecting anomalies of
APTs. The end result is detection output with high
resolution detecting anomalies. This process
illustrates how quantum machine learning can be used
for real time anomaly detection with data privacy.
Also, the research considers how quantum computing
affects encryption resilience. With quantum
simulators, the latest cryptography standards (RSA,
AES) are quantum attacked – as if they were post-
quantum. This assessment calls for Quantum-Safe
Key Exchange protocols proposed in the study to
secure cybersecurity resilience over the long term.
It’s a trial setting using open-source quantum
platforms like IBM Qiskit and Google Cirq to
simulate QSVM circuits. To check model
performance, we use benchmark datasets like
CICIDS2017 and NSL-KDD, accuracy, F1-score and
computational efficiency is used as tests.
4 EXPERIMENTAL RESULTS
AND DISCUSSION
The proposed QSVM was tested on encrypted
datasets to see whether it is effective in anomaly
detection. Results were compared with traditional
techniques, such as classic SVMs and deep learning
techniques, such as Convolutional Neural Networks
(CNN) and Long Short-Term Memory (LSTM)
networks. It revealed that QSVM had a much better
detection and computation performance, and beat
classical models for finding anomalies indicative of
APTs.
Figure 3: QSVM vs Classical Models - Radar Diagram.
Figure 3 show a radar graph shows the results of
QSVM and classic classical models (SVM, CNN,
LSTM) for three important performance indicators
(accuracy, precision and recall). The findings make it
clear that QSVM is more efficient than all classical
alternatives on every measure. QSVM achieves very
good precision and accuracy, whereas CNN and
LSTM fall behind because they can’t really extract
the data of encrypted communication with high
speed. This plot clearly shows that quantum kernels
have the edge over normalized kernels when it comes
Quantum Anomaly Detection for Advanced Persistent Threats Using Quantum Support Vector Machines
263

to discerning subtle trends in encrypted data, making
QSVM a better option for anomaly detection.
Figure 4: Quantum Attack Resilience of Encryption.
Figure 4 illustrates a heatmap plotting the strength
of current encryption protocols like RSA and AES for
different quantum circuit depths. These findings show
that RSA and AES encryption schemes lose their
security very quickly as quantum circuit complexity
increases, while the time required to decryption drops
significantly. On the contrary, the quantum-safe
encryption algorithms of this study remain much
more robust, even under very sophisticated quantum
attack simulations. The heatmap makes the point
visually in terms of the need to migrate to quantum-
resilient cryptography standards to protect data in the
quantum era.
Figure 5: QSVM vs Classical Models - Scalability
Comparison.
Figure 5 show the 3D surface plot QSVM versus
classical model scalability when the dataset size
increases linearly. This is shown by the graph which
shows that QSVM has a constant speed of processing,
which is very slow compared to classical models.
Classical models, on the other hand, are running at an
exponential rate in terms of computation time,
especially when the data base gets larger. These are
real-world examples of QSVM’s ability to detect
anomalies in real-time on large encrypted data sets,
and are therefore the ideal solution for solving high-
end cybersecurity challenges.
Table 1: Comparison of Performance Metrics.
Model Accuracy (%) Precision (%) Recall (%) F1-Score (%) False Positive Rate (%)
QSVM 97.8 98.1 97.5 97.8 2.1
Classical SVM 92.4 91 93.2 92.1 7.6
CNN 89.5 88.2 90 89.1 10.3
LSTM 90.2 89.8 90.5 90.1 9.8
Table 2: Computational Time Comparison Across Dataset Sizes.
Dataset Size QSVM (Time in s) Classical SVM (Time in s) CNN (Time in s) LSTM (Time in s)
1,000 2 5.5 7.8 6.3
5,000 5.1 15.2 18.4 17
10,000 10.5 30.1 38.2 36.3
50,000 20.8 65.4 85.6 80.5
1,00,000 35.4 120.2 140.8 135.6
Table 1 presents a detailed comparison of QSVM
and classical models (SVM, CNN, and LSTM) for
detecting anomalies, considering key performance
metrics: accuracy, precision, recall, F1-score, and
False Positive Rate (FPR). It is an extensive review
of each model strengths and weaknesses. The QSVM
outperforms all classical models in all performance
areas with highest accuracy, precision, F1-score and
lowest False Positive Rate. This is evidence of
QSVM’s success in picking out anomalies in
encrypted data, and reducing the false classifications.
Table 2 examines the computational efficiency of
ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,
COMMUNICATION, AND COMPUTING TECHNOLOGIES
264

QSVM compared to classical models when
processing datasets of increasing sizes. It indicates
the scaleability of every method, in seconds. QSVM
takes much less time to run as the dataset gets bigger
and works on a small size as well, scaling better than
classical models. QSVM’s processing time is linear
and in comparison, with classical approaches (SVM,
CNN, LSTM) it scales exponentially while traditional
approaches (SVM, CNN, LSTM) scale exponentially.
This indicates the scalability and practicality of
QSVM for real-time anomaly detection in big data.
Table 3 analyzes the time required to break
encryption standards RSA, AES, and the proposed
Quantum-Safe Encryption under varying levels of
quantum circuit depth. This highlights the resilience
of encryption algorithms to quantum attacks. The
results reveal that RSA and AES encryption standards
become highly vulnerable as quantum circuit depth
increases, with their resilience dropping drastically.
In contrast, the proposed Quantum-Safe Encryption
maintains robust resistance, highlighting its viability
as a secure alternative in a post-quantum
environment.
Table 3: Encryption Vulnerability Under Quantum Circuit Depth.
Quantum Circuit
Depth
RSA (Time to
Break, s)
AES (Time to
Break, s)
Quantum-Safe
(Time to Break,
s)
Vulnerability
Level
10 1,000 1,200 5,000 Low
20 500 700 4,800 Moderate
30 300 400 4,700 High
40 100 150 4,600 Very High
50 50 75 4,500 Critical
In addition, quantum simulators were used for
encryption resilience testing and RSA and AES
encryptions were discovered to be vulnerable to
quantum attacks. Although these standards of
encryption were protected from the old adversaries,
quantum decryption performed at much higher
efficiency. This shows the necessity of moving to
quantum resilient encryption algorithms. Algorithms
for Quantum-Safe Key Exchange had some
promising performance for data security in the face of
quantum threats in the future. This research is all
about the practical challenges of bringing quantum
computing into cybersecurity solutions. Detecting
APTs live on encrypted networks renders this work
an essential tool for the defense of valuable
infrastructure. Some limitations (like current
quantum hardware limitations) are acknowledged,
and suggestions for future work are to make QSVM
scalable for bigger datasets.
5 CONCLUSIONS
We present a novel quantum system for detecting
APTs using Quantum Support Vector Machines
(QSVM). Leveraging Quantum Kernel Methods, the
aforementioned model has higher accuracy and low
computational overhead in anomaly detection over
the classical ones. Quantum computing with real-time
decryption enabling real-time interpretation of
encrypted communication flows for privacy and
threat recognition accuracy. Besides, it points out
holes in current cryptographic standards in case of
quantum attacks and therefore recommends making
active efforts in [Q-Safe Key Exchange protocols].
Such results prove the transformative power of
quantum computing to drive cybersecurity resilience
against advanced adversaries such as APTs. This
work will also continue on scaling QSVM and
hardware implementations for quantum-enhanced
cybersecurity systems.
REFERENCES
C. Kenyon and C. Capano, “Apple silicon performance in
scientific computing,” in IEEE High Performance
Extreme Computing Conference (HPEC), 2022, pp. 1–
10.
D. Lakshmi, N. Nagpal, and S. Chandrasekaran, “A
quantum-based approach for offensive security against
cyber-attacks in electrical infrastructure,” Appl. Soft
Comput., vol. 136, p. 110071, 2023.
D. Said, “Quantum computing and machine learning for
cybersecurity: Distributed denial of service (DDoS)
attack detection on smart micro-grid,” Energies, vol.
16, no. 8, p. 3572, 2023.
E. F. Combarro, S. González-Castillo, and A. Di Meglio, A
Practical Guide to Quantum Machine Learning and
Quantum Optimization: Hands-on Approach to Modern
Quantum Algorithms. Packt Publishing Ltd., 2023.
Quantum Anomaly Detection for Advanced Persistent Threats Using Quantum Support Vector Machines
265

J. D. Bakos, Embedded systems: ARM programming and
optimization. Elsevier, 2023.
J. D. Bakos, Embedded Systems: ARM Programming and
Optimization. Elsevier, 2023.
K. Shara, “Quantum machine learning and cybersecurity,”
Quantum, vol. 12, no. 6, p. 47–56, 2023.
M. J. H. Faruk, A. Z. Mahmud, and I. Rahman, “A review
of quantum cybersecurity: threats, risks and
opportunities,” in IEEE International Conference on AI
in Cybersecurity (ICAIC), 2022, pp. 1–8.
M. J. H. Faruk, A. Z. Mahmud, and I. Rahman, “A review
of quantum cybersecurity: Threats, risks and
opportunities,” in IEEE International Conference on AI
in Cybersecurity (ICAIC), 2022, pp. 1–8.
M. Macas, C. Wu, and W. Fuertes, “A survey on deep
learning for cybersecurity: Progress, challenges, and
opportunities,” Comput. Netw., vol. 212, p. 109032,
2022.
M. S. Akter, I. R. Hossain, T. M. Ahmed, and S. M. Khan,
“Case study-based approach of quantum machine
learning in cybersecurity: Quantum support vector
machine for malware classification and protection,” in
IEEE 47th Annual Computers, Software and
Applications Conference (COMPSAC), 2023, pp.
1057–1063.
O. Faker and N. E. Cagiltay, “Quantum machine learning
in intrusion detection systems: A systematic mapping
study,” in International Conference on WorldS4,
Springer, 2023, pp. 99–113.
O. Faker and N. E. Cagiltay, “Quantum machine learning
in intrusion detection systems: A systematic mapping
study,” in International Conference on WorldS4,
Springer, 2023, pp. 99–113.
R. Kharsa, A. Bouridane, and A. Amira, “Advances in
quantum machine learning and deep learning for image
classification: A survey,” Neurocomputing, vol. 560, p.
126843, 2023.
R. Alluhaibi, “Quantum machine learning for advanced
threat detection in cybersecurity,” Int. J. Safety Security
Eng., vol. 14, no. 3, 2024.
R. Alluhaibi, “Quantum machine learning for advanced
threat detection in cybersecurity,” Int. J. Safety Security
Eng., vol. 14, no. 3, 2024.
S. K. Sood and M. Agrewal, “Quantum machine learning
for computational methods in engineering: A
systematic review,” Arch. Comput. Methods Eng., vol.
31, no. 3, pp. 1555–1577, 2024.
S. K. Sheoran and V. Yadav, “Comparative analysis of
classification efficiency of quantum machine learning
algorithms,” in IEEE International Conference on
Computing, Power and Communication Technologies
(IC2PCT), vol. 5, 2024, pp. 1818–1823.
W. S. Admass, Y. Y. Munaye, and A. A. Diro, “Cyber
security: state of the art, challenges and future
directions,” Cyber Secur. Appl., vol. 2, p. 100031,
2023.
Z. Ali, S. Kumar, and H. Ibrahim, “Reassessing the
performance of ARM vs x86 with recent technological
shift of Apple,” in IEEE International Conference on IT
and Industrial Technologies (ICIT), 2022, pp. 1–6.
ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,
COMMUNICATION, AND COMPUTING TECHNOLOGIES
266