Adversarial Deep Learning in Cybersecurity: Attack Taxonomies, Defence Mechanisms, and Learning Theories

A critical challenge in deep learning is the vulnerability of deep learning networks to security attacks from intelligent cyber adversaries. Even innocuous perturbations to the training data can be used to manipulate the behaviour of deep networks in unintended ways. In this book, we review the latest developments in adversarial attack technologies in computer vision; natural language processing; and cybersecurity with regard to multidimensional, textual and image data, sequence data, and temporal data. In turn, we assess the robustness properties of deep learning networks to produce a taxonomy of adversarial examples that characterises the security of learning systems using game theoretical adversarial deep learning algorithms. The state-of-the-art in adversarial perturbation-based privacy protection mechanisms is also reviewed.

We propose new adversary types for game theoretical objectives in non-stationary computational learning environments. Proper quantification of the hypothesis set in the decision problems of our research leads to various functional problems, oracular problems, sampling tasks, and optimization problems. We also address the defence mechanisms currently available for deep learning models deployed in real-world environments. The learning theories used in these defence mechanisms concern data representations, feature manipulations, misclassifications costs, sensitivity landscapes, distributional robustness, and complexity classes of the adversarial deep learning algorithms and their applications.

In closing, we propose future research directions in adversarial deep learning applications for resilient learning system design and review formalized learning assumptions concerning the attack surfaces and robustness characteristics of artificial intelligence applications so as to deconstruct the contemporary adversarial deep learning designs. Given its scope, the book will be of interest to Adversarial Machine Learning practitioners and Adversarial Artificial Intelligence researchers whose work involves the design and application of Adversarial Deep Learning.

深度學習面臨的一個重要挑戰是深度學習網絡對智能網絡攻擊的脆弱性。即使對訓練數據進行微小的干擾，也可以用來以意想不到的方式操縱深度網絡的行為。在本書中，我們回顧了計算機視覺、自然語言處理和多維、文本和圖像數據、序列數據和時間數據方面的對抗攻擊技術的最新發展。同時，我們評估了深度學習網絡的魯棒性特性，以建立一個以博弈論對抗深度學習算法為基礎的對抗性範例分類法，以表徵學習系統的安全性。同時，我們還回顧了基於對抗干擾的隱私保護機制的最新技術。

我們提出了新的對抗目標的對手類型，用於非穩態計算學習環境中的博弈論目標。對我們研究中的決策問題中的假設集進行適當的量化，導致了各種功能問題、神諭問題、抽樣任務和優化問題。我們還討論了目前在實際環境中部署的深度學習模型的防禦機制。這些防禦機制所使用的學習理論涉及數據表示、特徵操作、分類錯誤成本、敏感性地形、分布魯棒性以及對抗性深度學習算法及其應用的複雜性類。

最後，我們提出了對抗性深度學習應用於韌性學習系統設計的未來研究方向，並回顧了關於人工智能應用的攻擊面和魯棒性特徵的形式化學習假設，以解構當代對抗性深度學習設計。鑑於其範圍，本書將對從事對抗機器學習實踐和對抗人工智能研究的人員，特別是那些涉及對抗性深度學習的設計和應用的人員感興趣。

Dr. Aneesh Sreevallabh Chivukula is currently the Director of Artificial Intelligence at Adan Corporate. He has a PhD in data analytics and machine learning from the University of Technology Sydney (UTS), Australia. His research interests are in Computational Algorithms, Adversarial Learning, Intelligent Systems, Data Mining, and Data Science. He has been teaching subjects on advanced analytics and problem solving at UTS. He has industry experience in engineering, consulting, R&D at research labs and startup companies. He has developed enterprise solutions across the value chains in the open source, Cloud, & Big Data markets.

Dr. Xinghao Yang is currently an Associate Professor at the China University of Petroleum. He has a Ph.D. degree in advanced analytics from the University of Technology Sydney, Sydney, NSW, Australia. His research interests include multiview learning and adversarial machine learning with publications on information fusion and information sciences.

Dr. Wei Liu is the Director of Future Intelligence Research Lab, and an Associate Professor in Machine Learning, in the School of Computer Science, the University of Technology Sydney (UTS), Australia. He is a core member of the UTS Data Science Institute. Wei obtained his PhD degree in Machine Learning research at the University of Sydney (USyd). His current research focuses are adversarial machine learning, game theory, causal inference, multimodal learning, and natural language processing. Wei's research papers are constantly published in CORE A*/A and Q1 (i.e., top-prestigious) journals and conferences. He has received 3 Best Paper Awards. Besides, one of his first-authored papers received the Most Influential Paper Award in the CORE A Ranking conference PAKDD 2021. He was a nominee for the Australian NSW Premier's Prizes for Early Career Researcher Award in 2017. He has obtained more than $2 million government competitive and industry research funding in the past six years.

Dr. Bo Liu is currently a Senior Lecturer with the University of Technology Sydney, Australia. His research interests include cybersecurity and privacy, location privacy and image privacy, privacy protection and machine learning, wireless communications and networks. He is an IEEE Senior Member and Associate Editor of IEEE Transactions on Broadcasting.

Dr. Tianqing Zhu is an Associate Professor in Cyber Security in the Faculty of Engineering and IT at UTS, and the co-director of the Centre for Cyber Security & Privacy. She has extensive experience teaching and researching privacy preserving, cyber security and security in Artificial Intelligence. Tianqing's research has focused especially on differential privacy, an emerging model of cyber security that proponents claim can protect personal data far better than traditional methods. Tianqing is also interested in security and privacy in AI, including designing novel security models, developing efficient private algorithms, and performing in-depth analytics on a wide spectrum of AI areas.

Dr. Wanlei Zhou received the Ph.D. degree from Australian National University, Canberra, ACT, Australia, in 1991, all in computer science and engineering, and the D.Sc. degree from Deakin University, Melbourne, VIC, Australia, in 2002. He is currently a Professor and the Head of School of Computer Science at the University of Technology Sydney. He served as a Lecturer with the University of Electronic Science and Technology of China, a System Programmer with Hewlett Packard, Boston, MA, USA, and a Lecturer with Monash University, Melbourne, VIC, Australia, and the National University of Singapore, Singapore. He has published over 300 papers in refereed international journals and refereed international conferences proceedings. His research interests include distributed systems, network security, bioinformatics, and e-Learning. Dr. Wanlei was the General Chair/Program Committee Chair/Co-Chair of a number of international conferences, including ICA3PP, ICWL, PRDC, NSS, ICPAD, ICEUC, and HPCC.

Dr. Aneesh Sreevallabh Chivukula目前擔任Adan Corporate的人工智慧總監。他在澳大利亞悉尼科技大學(UTS)獲得數據分析和機器學習的博士學位。他的研究興趣包括計算算法、對抗學習、智能系統、數據挖掘和數據科學。他在UTS教授高級分析和問題解決課程。他在工程、咨詢、研究實驗室和初創公司的行業經驗。他在開源、雲和大數據市場上開發了企業解決方案。

Dr. Xinghao Yang目前是中國石油大學的副教授。他在澳大利亞悉尼科技大學獲得高級分析的博士學位。他的研究興趣包括多視圖學習和對抗機器學習，並在信息融合和信息科學方面發表了論文。

Dr. Wei Liu是未來智能研究實驗室的主任，也是澳大利亞悉尼科技大學計算機科學學院的副教授。他是悉尼大學機器學習研究的核心成員。Wei在悉尼大學獲得機器學習研究的博士學位。他目前的研究重點是對抗機器學習、博弈論、因果推斷、多模態學習和自然語言處理。Wei的研究論文經常發表在CORE A*/A和Q1（即頂級）期刊和會議上。他獲得了3個最佳論文獎。此外，他的一篇第一作者論文在CORE A Ranking會議PAKDD 2021中獲得了最具影響力論文獎。他曾經入圍2017年澳大利亞新南威爾士州總理早期職業研究員獎。在過去六年中，他獲得了超過200萬美元的政府競爭性和行業研究資金。

Dr. Bo Liu目前是澳大利亞悉尼科技大學的高級講師。他的研究興趣包括網絡安全和隱私、位置隱私和圖像隱私、隱私保護和機器學習、無線通信和網絡。他是IEEE的高級會員和IEEE Transactions on Broadcasting的副編輯。

Dr. Tianqing Zhu是悉尼科技大學工程與信息技術學院的網絡安全副教授，也是Cyber Security & Privacy中心的聯合主任。她在隱私保護、網絡安全和人工智慧安全方面有豐富的教學和研究經驗。Tianqing的研究尤其關注差分隱私，這是一種新興的網絡安全模型，據稱可以比傳統方法更好地保護個人數據。Tianqing還對人工智慧中的安全和隱私感興趣，包括設計新的安全模型、開發高效的私有算法，以及對各種人工智慧領域進行深入分析。

Dr. Wanlei Zhou在1991年從澳大利亞國立大學獲得計算機科學和工程博士學位，並在2002年從迪肯大學獲得D.Sc.學位。他目前是悉尼科技大學計算機科學學院的教授和學院主任。他曾在中國電子科技大學擔任講師，並在美國波士頓的惠普公司和澳大利亞的蒙納士大學和新加坡國立大學擔任系統程序員和講師。他在國際期刊和會議上發表了300多篇論文。他的研究興趣包括分布式系統、網絡安全和生物信息學。