Transfer learning provides an effective solution for feasibly and fast customize accurate Student models, by transferring the learned knowledge of pre-trained Teacher models over large datasets via fine-tuning. Many pre-trained Teacher models used in transfer learning are publicly available and maintained by public platforms, increasing their vulnerability to backdoor attacks. In this paper, we demonstrate a backdoor threat to transfer learning tasks on both image and time-series data leveraging the knowledge of publicly accessible Teacher models, aimed at defeating three commonly-adopted defenses: pruning-based, retraining-based and input pre-processing-based defenses. Specifically, (A) ranking-based selection mechanism to speed up the backdoor trigger generation and perturbation process while defeating pruning/retraining-based defenses. (B) autoencoder-powered trigger generation is proposed to produce a robust trigger that can defeat the input pre-processing-based defense. (C) defense-aware retraining to generate the manipulated model. We launch effective misclassification attacks on Student models over real-world images, brain MRI and ECG data. Experiments reveal that our attack can maintain the 98.4% and 97.2% classification accuracy as the genuine model on clean image and time series inputs respectively while improving 27.9%-100% and 27.1% -56.1% attack success rate on trojaned image and time series inputs respectively in the presence of pruning-based and/or retraining-based defenses.

Wearable fitness devices are widely used to track an individual’s health and physical activities to improve the quality of health services. These devices sense a considerable amount of sensitive data processed by a centralized third party. While many researchers have thoroughly evaluated privacy issues surrounding wearable fitness trackers, no study has addressed privacy issues in trackers by giving control of the data to the user. Blockchain is an emerging technology with outstanding advantages in resolving consent management privacy concerns. As there are no fully transparent, legally compliant solutions for sharing personal fitness data, this study introduces an architecture for a human-centric, legally compliant, decentralized and dynamic consent system based on blockchain and smart contracts. Algorithms and sequence diagrams of the proposed system’s activities show consent-related data flow among various agents, which are used later to prove the system’s trustworthiness by formalizing the security requirements. The security properties of the proposed system were evaluated using the formal security modeling framework SeMF, which demonstrates the feasibility of the solution at an abstract level based on formal language theory. As a result, we have empirically proven that blockchain technology is suitable for mitigating the privacy issues of fitness providers by recording individuals’ consent using blockchain and smart contracts.

Due to the increasing integration of distributed energy generation in the electric grid, transactive energy markets (TEMs) have recently emerged to balance the demand and supply dynamically across the grid. TEM enables peer to peer (P2P) energy trading and brings flexibility by reducing users’ demand in the grid. It also enhances the system’s efficiency and reduces the pressure on electricity networks. However, it is vulnerable to major cyber attacks as users equipped with smart devices are participating autonomously in the energy market, and an extensive amount of information is exchanged through the communication channel. The potential attacks and impacts of those attacks need to be investigated to develop an attack resilient TEM-based power system. Hence, in this paper, our goal is to systematically identify possible cyber attacks associated with a TEM-based power system. In order to achieve this goal, we classify the attacks during the P2P and flexibility schemes of TEM into three main categories. Then, we explore the attacks under each category in detail. We further distinguish the adversary roles of each particular attack and see what benefits will be received by an adversary through each specific attack. Finally, we present the impact of the attacks on the market operation, consumers, and prosumers of the TEM in this paper.

The susceptibility of deep neural networks (DNNs) to adversarial intrusions, exemplified by adversarial examples, is well-documented. Conventional attacks implement unstructured, pixel-wise perturbations to mislead classifiers, which often results in a noticeable departure from natural samples and lacks human-perceptible interpretability. In this work, we present an adversarial attack strategy that implements fine-granularity, semantic-meaning-oriented structural perturbations. Our proposed methodology manipulates the semantic attributes of images through the use of disentangled latent codes. We engineer adversarial perturbations by manipulating either a single latent code or a combination thereof. To this end, we propose two unsupervised semantic manipulation strategies: one based on vector-disentangled representation and the other on feature map-disentangled representation, taking into consideration the complexity of the latent codes and the smoothness of the reconstructed images. Our empirical evaluations, conducted extensively on real-world image data, showcase the potency of our attacks, particularly against black-box classifiers. Furthermore, we establish the existence of a universal semantic adversarial example that is agnostic to specific images.