ExplainabilityTransparencySecuritycs.CV
Explainable Artificial Intelligence to Detect Image Spam Using Convolutional Neural Network
Zhibo Zhang, Ernesto Damiani, Hussam Al Hamadi, Chan Yeob Yeun, Fatma Taher
September 7, 2022
International Conference Control and Robots
11 citations
Key Takeaways
-
CNN models successfully detected spam images with satisfactory performance across a dataset of 6,636 images from real email sources, demonstrating that automated systems can effectively identify this type of threat that poses security risks as internet usage expands.
-
Explainability methods (LIME and SHAP) made the AI decision-making process transparent by revealing which visual features the spam detection model relied on to make its classifications, addressing the "black box" problem where users cannot understand why an AI system made a particular decision.
-
Model-independent explanation tools can work across different AI systems, meaning organizations don't need to rebuild their explainability infrastructure if they switch spam detection models—the same XAI methods can provide insights into different underlying technologies.
-
Transparency in spam detection has practical security benefits: by understanding which image characteristics trigger spam alerts, security teams can better evaluate whether the system is making decisions based on legitimate spam indicators or potentially problematic patterns, supporting more responsible deployment of these systems.
Abstract
Image spam threat detection has continually been a popular area of research with the internet's phenomenal expansion. This research presents an explainable framework for detecting spam images using Convolutional Neural Network(CNN) algorithms and Explainable Artificial Intelligence (XAI) algorithms. In this work, we use CNN model to classify image spam respectively whereas the post-hoc XAI methods including Local Interpretable Model Agnostic Explanation (LIME) and Shapley Additive Explanations (SHAP) were deployed to provide explanations for the decisions that the black-box CNN models made about spam image detection. We train and then evaluate the performance of the proposed approach on a 6636 image dataset including spam images and normal images collected from three different publicly available email corpora. The experimental results show that the proposed framework achieved satisfactory detection results in terms of different performance metrics whereas the model-independent XAI algorithms could provide explanations for the decisions of different models which could be utilized for comparison for the future study.