Special Session Title: Generative Adversarial Attacks and Defence Strategies Against Deep Networks
Institution Name and Country: University of Liverpool/AURAK, UK, UAE.
Session Chair/s (Title, Name): Dr Ali Al-Ataby
Session Co-Chair/s (If Applicable): Prof Fawzi Al-Naima
Session Committee: TBC
Synopsis:
The widespread usage of machine and deep learning tools in our daily life raises concerns over the security, reliability and robustness of such tools. Adversarial machine/deep learning is a technique that attempts to exploit a machine learning model by examining decision boundaries. It is a very active research area which involves developing attack mechanisms to trick existing models or defence strategies to detect or stop such attacks. In this case, carefully constructed examples that cause machine learning models to provide unexpected outputs are referred to as adversarial samples. For the machine learning systems, these samples can be regarded as visual illusions that could manipulate the behaviour and output of these intelligent tools.
Topics:
This special session invites authors to submit high-quality research papers on emerging technologies with regards to adversarial attacks and defence strategies, covering topics which include (but are not limited to) the following:
- Exploring adversarial samples
- Classifying and characterising adversarial attacks and defence strategies against deep neural networks
- Evaluating and measuring the robustness of deep neural networks against adversarial examples
- Investigating transferability of adversarial examples among different system setups (e.g. black box, while box).
- Exploring adversarial attacks against machine and deep learning-based intrusion detection systems (IDS) and other network security tools
My info:
Senior member of IEEE (SMIEE)
Membership no. 98637215