Deep Learning Attacks and Active Defences: A Cybersecurity Perspective

点击翻译按钮获取中文摘要

The belief that deep learning technology is imperative for economic development, military control, and strategic competitiveness has accelerated its development across the globe. However, experience has revealed the disappointing fact that deep learning models are vulnerable to a range of security attacks. Hence, a series of methodologies and defence strategies will be devised that make deep learning systems robust to these attacks. The methodologies require analysing attack lifecycles to identify them in their early stages. With this knowledge, active defence methods and forensic strategies can be developed to ensure efficient defences and prevent further attacks. Moreover, the outputs will be generalisable to most deep learning services.