Jin Li, E-mail: email@example.com (Managing Guest Editor)
Guangzhou University, Guangzhou, China
Peng Cheng Laboratory, Shenzhen, China
Witold Pedrycz, E-mail: firstname.lastname@example.org
University of Alberta, Edmonton, AB, Canada.
Changyu Dong, E-mail: email@example.com
Newcastle University, UK
With the rapid development of data science, machine learning has been widely applied to many important fields such as computer vision, healthcare systems, and financial predictions, to support the design of constructs of Artificial Intelligence.
However, the environment of AI and machine learning is adversarial, in which machine learning tasks are faced with a variety of essential security threats coming from multiple parties. In fact, a malicious adversary can carefully manipulate input data, learning procedures, and outputs by exploiting specific vulnerabilities of learning tasks to compromise the security of machine learning systems. In terms of adversaries’ goals, the threats can be divided into three main categories: security violation, attack specificity, and error specificity. Examples include: 1) aiming to evade detection without compromising normal system operation; 2) aiming to cause misclassification of a specific set of samples, or of any sample; 3) aiming to have a sample misclassified as a specific class, or any of the classes different from the true class.
To understand the security properties of Adversarial Machine Learning (AML), one should address the following main issues: 1) Identifying potential vulnerabilities of machine learning algorithms during learning, and classification; 2) Devising appropriate attacks that correspond to the identified threats and evaluating their impact on the targeted system; 3) Proposing countermeasures to improve the security of machine learning algorithms against the considered attacks.
This feature topic will benefit the research community towards identifying challenges and disseminating the latest methodologies and solutions to adversarial machine learning. The ultimate objective is to publish high-quality articles presenting open issues, delivering algorithms, protocols, frameworks, and solutions. All received submissions will be sent out for peer review by at least three experts in the field and evaluated with respect to relevance to the special section, level of innovation, depth of contributions, and quality of presentation. Case studies, which address state-of-art research and state-of-practice industry experiences, are also welcome. Guest Editors will make an initial determination of the suitability and scope of all submissions. Papers that either lack originality, clarity in presentation or fall outside the scope of the special issue will not be sent for review and the authors will be promptly notified in such cases. Submitted papers must not be under consideration by any other journal or publication
Topics of interest include, but are not limited to, the following:
Dependable Machine Learning Algorithms in Adversarial Setting
Secure Federated Machine Learning against Malicious Attacker
Security Evaluation for Federated Machine Learning
Privacy Disclosure in Traditional Machine Learning Algorithms
Adversarial Sample and Its Detection Method for Natural Language Processing
Adversarial Sample and Its Detection Method for Images Recognition
Malware Code Manipulation against Detection based on Machine Learning
Verification Mechanism for Data-outsourced Machine Learning
Explanation Methods for Machine Learning Application
Submission Starting: Feb. 20th, 2020
Notification Date: 3-4 months
Submission Closing: Oct. 20th, 2020
We will broadcast the SI CFP information in many ways, including the Applied Soft Computing website and mail lists etc. The number of submissions is expected over 100 submissions. We plan to accept around 15-25 finally, with acceptance rate 15%.
Highlight of the Topic
Machine-learning-based artificial intelligent applications have been shown that adversarial input perturbations carefully crafted either at training or at test time can easily subvert their predictions, which certainly poses security threats to the applications. These include, e.g., manipulation of malware code to have the corresponding sample misclassified as legitimate, or manipulation of images to mislead object recognition. The vulnerability of machine learning in such an adversarial environment, along with the design of suitable countermeasures, have been investigated in the research field of adversarial machine learning. Therefore, this particular topic will benefit the readers towards identifying challenges and disseminating the latest methodologies and solutions to adversarial machine learning.
Paper submissions for this special issue should follow the submission format and guidelines for regular papers at the Applied Soft Computing website: http://www.journals.elsevier.com/applied-soft-computing/. Authors’ Manuscript should be submitted online at EES system: http://ees.elsevier.com/asoc/. Please select the special issue name in EES system with the article type as “SI: Techniques in Adversarial ML”.