The ability to perform mobile traffic forecasting effectively with Deep Neural Networks (DNN) is instrumental to optimise resource management in 5G and beyond generation mobile networks. However, despite their capabilities, these DNNs often act as complex opaque-boxes with decisions that are difficult to interpret. Even worse, they have proven vulnerable to adversarial attacks which undermine their applicability in production networks. Unfortunately, the existing state of the art EXplainable Artificial Intelligence (XAI) techniques which are primarily tailored for computer vision and Natural Language Processing (NLP) fall short when applied to spatio-temporal time series forecasting models. To address these challenges, we introduce DEEXP, a tool that flexibly builds upon legacy XAI techniques to synthesise semantically compact explanations by making it possible to understand which Base Stations (BS)s are more influential for forecasting from a spatio-temporal perspective. Armed with such knowledge, we run state-of-the art AML techniques on those BSs to measure the accuracy degradation of the predictors. Our comprehensive evaluation uses real-world mobile traffic datasets and demonstrates that legacy XAI techniques spot different types of vulnerabilities. While Gradient-weighted Class Activation Mapping (GC) is suitable to spot BSs sensitive to moderate/low traffic injection, LayeR-wise backPropagation (LRP) is suitable to identify BSs sensitive to high traffic injection. Under moderate/subtle adversarial attacks, the prediction error can increase by more than 400%.

About Serly Moghadas

Serly is a PhD student in Wireless Networking Group (WNG) under the supervision of Prof. Joerg Widmer. She obtained her MSc degree in Telecommunications Engineering from Urmia University in 2018. Her research focuses on developing explainable AI tools for mobile networks and exploring the intersection of machine learning with mobile networks.

This event will be conducted in English