Abstract

QR codes are extensively used for connecting physical items with online platforms. However, this ease of QR codes can also be exploited in quishing attacks resulting in the QR code leading Redirects the victim to a harmful site. A malicious labeler could be one who only marks a QR code as malicious. Hard for a user to believe, since there is no visible reason to the user for the decision. The paper introduces a conceptual model for detecting QR code phishing which uses the concept of machine learning classifiers and XAI. The model features a QR code payload extraction step and it outputs a suitable QR code. Has QR code features at the QR code level. To explain global SHapley Additive exPlanations (SHAP) are suggested. LIME generates feature attribution that is interpretable and model-agnostic.and instance-level feature attribution, is model agnostic and interpretable. As an alternative to the local explanation method, (LIME) can be used. The framework also is translated. technical feature contributions into brief, user friendly warning messages. The paper does not present experimental results but outlines a reproducible evaluation strategy that includes The following metrics: precision, recall, F1 SCORE, ROC AUC, false positive rate, explanation consistency, and inference speed. The main contribution is a design that makes explainability a part of every design. As an integral component of the detection process, not an optional report that is generated after classification. Moreover, the paper touches upon the issues of dataset bias, phishing behaviour changes, feature Attribution limitations, privacy, and implementation are all issues. The term “quishing” is a portmanteau of the words “quick” and “phishing” that refers to the practice of exploiting QR codes for phishing attacks.The term “quishing” is derived from “quick” and “phishing” and it describes the exploitation of QR codes in phishing attacks. SHAP, LIME, cybersecurity.

Keywords

How to Cite This Article

Rights