Preface of the 2018 Symposium on Adversary Aware Learning Techniques and Trends in Cybersecurity (ALEC) (co-located with AAAI Fall Symposium Series 2018) Prithviraj Dasgupta gupta@unomaha.edu 0 1 2 3 4 5 6 7 Joseph B. Collins 0 1 2 3 4 5 6 7 Ranjeev Mittu ranjeev.mittug@nrl.navy.mil 0 1 2 3 4 5 6 7 Amitabh Mishra 0 1 2 3 4 5 6 7 U.S. Army CERDEC 0 1 2 3 4 5 6 7 Abebaw Tadesse, Langston University , USA Adversary-aware Machine Learning - Reinforcement Learning , Lifelong Learning, Deep Learning Adversaryaware Prediction , Forecasting and Decision Making Techniques Applications of Adversarial Learning Krishnendu Ghosh, Miami University of Ohio , USA Operations Research related to Adversarial Learning P. Dasgupta is with the Computer Science Department, University of Nebraska , Omaha, NE , USA Ying Zhao, Naval Postgraduate School , USA

Machine learning-based intelligent systems have experienced a massive growth over the past few years, and are close to becoming ubiquitous in the technology surrounding our daily lives. However, a critical challenge in machine learning-based systems is their vulnerability to security attacks from malicious adversaries. The vulnerability of these systems is further aggravated as it is non-trivial to establish the authenticity of data used to train the system, and even innocuous perturbations to the training data can be used to manipulate the systems behavior in unintended ways. Recent reports about malicious manipulation of social media feeds masquerading as authentic news items provide compelling evidence towards developing stronger and more resilient measures for combating adversarial attacks on machine learning-based systems. The ALEC'18 symposium was organized to address the overarching need towards making automated, machine learning-based systems more robust and resilient against adversarial attacks, so that humans can use them in a safe and sustained manner. The areas of interest of the symposium included the following topics: Generative Adversarial Networks Security Threats and Vulnerabilities of Adversarial Learning

 -

Human factors and adversarial learning with human-inthe-loop

The symposium included two keynote talks and ten orally presented papers. The first keynote talk titled AI Canonical Architecture and Robust AI by David R. Martinez from MIT Lincoln Laboratories discussed the performance assessment of AI-based systems and the need for robust AI. Xiaojin (Jerry) Zhu from the University of Wisconsin-Madison presented the second keynote titled An Optimal Control View of Adversarial Machine Learning on a novel control theorybased framework for representing various adversarial learning problems. The research papers presented at the symposium were grouped into three theme-based sessions: (1) Adversarial Data Generation and Adversarial Training, (2) Countering Adversarial Attacks in Cybersecurity, and, (3) Novel Approaches in Adversarial Artificial Intelligence. The symposium concluded with a group discussion on the immediate technological enablers and hurdles in adversarial learning as well as determining a roadmap for addressing longer term problems and challenges in the field.

Finally, we would like to thank the following ALEC’18 program committee members and reviewers for their support with reviewing papers and with various aspects of organizing the symposium: