Home / Quantum / FYP

Quantum Computing & Final-Year Research

My BS Computer Science capstone: can a quantum support vector machine beat a classical one at detecting spam email?

Final-Year ProjectQSVMQiskitPCAResearch
§ 01Final-Year Project

QSVM vs Classical SVM for Email-Spam Detection.

A controlled comparison of quantum and classical kernel methods on the same classification task.

FYP Quantum Machine Learning · Research

The research question

Quantum machine learning promises richer feature spaces through quantum kernels — but does that translate into measurably better classification on a real, messy security problem? I built both a Quantum SVM (QSVM) and a Classical SVM for email-spam detection and compared them under identical data conditions.

The constraint that shapes everythingCurrent quantum hardware and simulators support only a small number of qubits, so high-dimensional text features can't be encoded directly. PCA becomes essential — reducing features to fit the qubit budget while preserving as much signal as possible.
PythonQiskitQSVMQuantum KernelsScikit-LearnPCANLTKTF-IDF
QSVM-vs-SVM-for-Email-Spam-Detection ↗

01 Methodology

  • Email corpus preprocessing — tokenisation, stop-word removal, and vectorisation of message text.
  • PCA dimensionality reduction to compress features into a qubit-feasible space.
  • Quantum feature mapping — encoding classical data into quantum states to compute a quantum kernel.
  • Classical SVM trained on identical reduced features as the control.
  • Comparison across accuracy, precision, recall, and training cost.

02 Why it matters for security

  • Spam and phishing detection is a frontline security control — small accuracy gains matter at scale.
  • Tests whether quantum kernels offer real advantage or just novelty on practical data.
  • Probes where AI reliability breaks down in threat-detection contexts.
  • Builds intuition for post-quantum thinking in security architecture.

03 What I took away

The interesting result isn't a headline number — it's understanding the trade-off surface: where PCA compression costs more accuracy than the quantum kernel gains back, how simulator limitations bound the experiment, and why honest baselines matter more than impressive claims. The same discipline shows up in my threat-intel work, where I caught a data-leakage artefact inflating an F1 score from 0.27 to 0.90.

§ 02Foundations

Quantum computing coursework.

The theory behind the project.

Module 01

Mathematical Foundations & Quantum Mechanics

Complex numbers, matrix operations, and linear transformations — the mathematics underpinning qubits, superposition, and entanglement.

Module 02

Python Programming for Quantum Computing

Building quantum circuits and running them on simulators and real quantum hardware using Python and Qiskit.

Module 03

Qiskit & Advanced Algorithms

Quantum algorithms including Bernstein-Vazirani, Grover's, and Shor's — both the theory and hands-on implementation.

§ 03Credentials

Quantum certification.

Click the certificate name to view it.

Packt · Coursera Specialization (3 courses) · Nov 16, 2025
⤓ PDF
AVAILABLE FOR SOC · RED TEAM · NETWORK ROLES

Interested in the research?

Happy to discuss the methodology, results, or quantum machine learning generally.

Email me Résumé ↓ LinkedIn ↗
Emailshahyankhan886@gmail.com Phone+92 310 5945021 GitHubgithub.com/Asim223 LinkedInhafiz-asim-shayan-khan