Bridging Human and Machine Judgment : improving phishing detection through comparative analysis

Read full text Summarise with VIRAT
Authors: Kristine Madugay
Cited by: 0
References: 7

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Phishing attacks exploit human trust through psychological manipulation, and automated detection systems struggle with context-dependent indicators due to rigid binary classification. In collaboration with Phisherie Oy, this thesis aims to provide guidance in restructuring Phisherie's 29-criteria model by applying a multi-point scoring approach that replaces binary classification. This enables testing the reliability of each indicator in signaling phishing attempts while exploring opportunities to improve contextual reasoning in phishing detection. Using a controlled dataset of eight emails in a mixed-method approach, the experimental analysis revealed systematic weaknesses in automated systems due to binary limitations, while the human-based multi-point system correctly classified all eight emails through contextual human judgment. A comparative analysis of the three detection approaches (heuristic rule-based, AI-based, and human-based) identified distinct performance gaps and transferable strengths. The findings revealed that automation carries inherent risks; consequently, human-in-the-loop (HITL) systems are a recommended approach. HITL not only combines automated analysis with human judgment to effectively mitigate automation vulnerabilities through contextual reasoning, but also reinforces continuous learning through transparency, enabling users to improve their detection judgment over time. These results provide evidence-based guidance for enhancing phishing detection and enabling more informed cybersecurity decision-making.

Fields of Study & Subjects

Spam and Phishing Detection User Authentication and Security Systems Misinformation and Its Impacts

Publication details

Journal / source
Theseus (Ammattikorkeakoulujen)
Publication date
1 January 2026
Language
EN
Document type
Thesis