Biblio

Artificial Intelligence also often referred to as machine learning is being labelled to as the future has been into light since more than a decade. Artificial Intelligence designated by the acronym AI has a vast scope of development and the developers have been working on with it constantly. AI is being associated with the existing objects in the world as well as with the ones that are about to arrive to improve them and make them more reliable. AI as it states in its name is intelligence, intelligence shown by the machines to work similar to humans and work on achieving the goals they are being provided with. Another application of AI could be to provide defenses against the present cyber threats, vehicle overrides etc. Also, AI might be intelligence but, in the end, it's still a bunch of codes, hence it is prone to be corrupted or misused by the world. To prevent the misuse of the technologies, it is necessary to deploy them with a sustainable defensive system as well. Obviously, there is going to be a default defense system but it is prone to be corrupted by the hackers or malfunctioning of the intelligence in certain scenarios which can result disastrous especially in case of Robotics. A proposal referred to as the “Guard Masking” has been offered in the following paper, to provide an alternative for securing Artificial Intelligence.

Advances in deep neural networks (DNNs) have shown tremendous promise in the medical domain. However, the deep learning tools that are helping the domain, can also be used against it. Given the prevalence of fraud in the healthcare domain, it is important to consider the adversarial use of DNNs in manipulating sensitive data that is crucial to patient healthcare. In this work, we present the design and implementation of a DNN-based image translation attack on biomedical imagery. More specifically, we propose Jekyll, a neural style transfer framework that takes as input a biomedical image of a patient and translates it to a new image that indicates an attacker-chosen disease condition. The potential for fraudulent claims based on such generated `fake' medical images is significant, and we demonstrate successful attacks on both X-rays and retinal fundus image modalities. We show that these attacks manage to mislead both medical professionals and algorithmic detection schemes. Lastly, we also investigate defensive measures based on machine learning to detect images generated by Jekyll.

Following the recent adoption of deep neural networks (DNN) accross a wide range of applications, adversarial attacks against these models have proven to be an indisputable threat. Adversarial samples are crafted with a deliberate intention of undermining a system. In the case of DNNs, the lack of better understanding of their working has prevented the development of efficient defenses. In this paper, we propose a new defense method based on practical observations which is easy to integrate into models and performs better than state-of-the-art defenses. Our proposed solution is meant to reinforce the structure of a DNN, making its prediction more stable and less likely to be fooled by adversarial samples. We conduct an extensive experimental study proving the efficiency of our method against multiple attacks, comparing it to numerous defenses, both in white-box and black-box setups. Additionally, the implementation of our method brings almost no overhead to the training procedure, while maintaining the prediction performance of the original model on clean samples.

Video surveillance, closed-circuit TV and IP-camera systems became virtually omnipresent and indispensable for many organizations, businesses, and users. Their main purpose is to provide physical security, increase safety, and prevent crime. They also became increasingly complex, comprising many communication means, embedded hardware and non-trivial firmware. However, most research to date focused mainly on the privacy aspects of such systems, and did not fully address their issues related to cyber-security in general, and visual layer (i.e., imagery semantics) attacks in particular. In this paper, we conduct a systematic review of existing and novel threats in video surveillance, closed-circuit TV and IP-camera systems based on publicly available data. The insights can then be used to better understand and identify the security and the privacy risks associated with the development, deployment and use of these systems. We study existing and novel threats, along with their existing or possible countermeasures, and summarize this knowledge into a comprehensive table that can be used in a practical way as a security checklist when assessing cyber-security level of existing or new CCTV designs and deployments. We also provide a set of recommendations and mitigations that can help improve the security and privacy levels provided by the hardware, the firmware, the network communications and the operation of video surveillance systems. We hope the findings in this paper will provide a valuable knowledge of the threat landscape that such systems are exposed to, as well as promote further research and widen the scope of this field beyond its current boundaries.