Towards Robustifying Image Classifiers against the Perils of Adversarial Attacks on Artificial Intelligence Systems.

Adversarial machine learning (AML) is a class of data manipulation techniques that cause alterations in the behavior of artificial intelligence (AI) systems while going unnoticed by humans. These alterations can cause serious vulnerabilities to mission-critical AI-enabled applications. This work introduces an AI architecture augmented with adversarial examples and defense algorithms to safeguard, secure, and make more reliable AI systems. This can be conducted by robustifying deep neural network (DNN) classifiers and explicitly focusing on the specific case of convolutional neural networks (CNNs) used in non-trivial manufacturing environments prone to noise, vibrations, and errors when capturing and transferring data. The proposed architecture enables the imitation of the interplay between the attacker and a defender based on the deployment and cross-evaluation of adversarial and defense strategies. The AI architecture enables (i) the creation and usage of adversarial examples in the training process, which robustify the accuracy of CNNs, (ii) the evaluation of defense algorithms to recover the classifiers' accuracy, and (iii) the provision of a multiclass discriminator to distinguish and report on non-attacked and attacked data. The experimental results show promising results in a hybrid solution combining the defense algorithms and the multiclass discriminator in an effort to revitalize the attacked base models and robustify the DNN classifiers. The proposed architecture is ratified in the context of a real manufacturing environment utilizing datasets stemming from the actual production lines.

Automated Cyber and Privacy Risk Management Toolkit.

Addressing cyber and privacy risks has never been more critical for organisations. While a number of risk assessment methodologies and software tools are available, it is most often the case that one must, at least, integrate them into a holistic approach that combines several appropriate risk sources as input to risk mitigation tools. In addition, cyber risk assessment primarily investigates cyber risks as the consequence of vulnerabilities and threats that threaten assets of the investigated infrastructure. In fact, cyber risk assessment is decoupled from privacy impact assessment, which aims to detect privacy-specific threats and assess the degree of compliance with data protection legislation. Furthermore, a Privacy Impact Assessment (PIA) is conducted in a proactive manner during the design phase of a system, combining processing activities and their inter-dependencies with assets, vulnerabilities, real-time threats and Personally Identifiable Information (PII) that may occur during the dynamic life-cycle of systems. In this paper, we propose a cyber and privacy risk management toolkit, called AMBIENT (Automated Cyber and Privacy Risk Management Toolkit) that addresses the above challenges by implementing and integrating three distinct software tools. AMBIENT not only assesses cyber and privacy risks in a thorough and automated manner but it also offers decision-support capabilities, to recommend optimal safeguards using the well-known repository of the Center for Internet Security (CIS) Controls. To the best of our knowledge, AMBIENT is the first toolkit in the academic literature that brings together the aforementioned capabilities. To demonstrate its use, we have created a case scenario based on information about cyber attacks we have received from a healthcare organisation, as a reference sector that faces critical cyber and privacy threats.