Smishlock Holmes: Explainable Smishing Detector

Oppenheimer Marques, Andre Luis (2025) Smishlock Holmes: Explainable Smishing Detector. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (4MB) | Preview
Preview	PDF (Configuration Manual) Download (4kB) | Preview

Abstract

Smishing (SMS phishing) is a rapidly evolving social engineering threat that exploits both technical evasion tactics and cognitive vulnerabilities in mobile-first environments. Traditional defences often fail against adversarially crafted messages and rarely provide explanations that enhance user awareness and trust. This study investigates how Large Language Models (LLMs) can be applied to detect smishing messages in a way that combines classification robustness with interpretable, context-aware explanations for cognitively constrained mobile users.

We present Smishlock Holmes, a modular AI framework that detects technical, psychological, and exploitation cues, normalises obfuscated text, and produces causal narratives linking cues to risk. Built on a JSON-Embedded Few-Shot Chain-of-Thought architecture, it enforces schema-validated, model-agnostic outputs.

Using a balanced subset of the Super SMS Dataset, classification accuracy and explanation quality were assessed, the latter via the G-Eval framework. The system achieved 87.5% accuracy, with high correctness (4.68/5) and clarity (4.50/5), but lower completeness (3.94/5).

These results show that LLM-based detection can unify adversarial resilience and interpretability in a single pipeline. Future work will expand cue coverage, adopt multiagent verification, and explore privacy-preserving on-device inference for scalable, user-adaptive smishing defences.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Aleburu, Joel UNSPECIFIED
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science P Language and Literature > P Philology. Linguistics > Computational linguistics. Natural language processing Q Science > QA Mathematics > Computer software > Computer Security T Technology > T Technology (General) > Information Technology > Computer software > Computer Security
Divisions:	School of Computing > Master of Science in Cyber Security
Depositing User:	Ciara O'Brien
Date Deposited:	16 Jun 2026 14:12
Last Modified:	16 Jun 2026 14:12
URI:	https://norma.ncirl.ie/id/eprint/9367

Actions (login required)

View Item