Search

While 5G can facilitate high-speed Internet access and make over-the-horizon control a reality for unmanned aerial vehicles (UAVs; also known as drones), there are also potential security and privacy considerations, for example, authentication among drones. Centralized authentication approaches not only suffer from a single point of failure, but they are also incapable of cross-domain authentication, which complicates the cooperation of drones from different domains. To address these limitations, a blockchain-based cross-domain authentication scheme for intelligent 5G-enabled Internet of drones is proposed in this paper. Our approach employs multiple signatures based on threshold sharing to build an identity federation for collaborative domains. This allows us to support domain joining and exiting. Reliable communication between cross-domain devices is achieved by utilizing smart contract for authentication. The session keys are negotiated to secure subsequent communication between two parties. Our security and performance evaluations show that the proposed scheme is resistant to common attacks targeting Internet of Things (IoT) devices (including drones), as well as demonstrating its effectiveness and efficiency.

Due to its nonintrusive character, WiFi channel state information (CSI)-based activity recognition has attracted tremendous attention in recent years. Since activity recognition performance heavily relies on activity segmentation results, a number of activity segmentation methods have been designed, and most of them focus on seeking optimal thresholds to segment activities. However, these threshold-based methods are strongly dependent on designers’ experience and might suffer from performance decline when applying to the scenario, including both fine-grained and coarse-grained activities. To address these challenges, we present DeepSeg, a deep learning-based activity segmentation framework for activity recognition using WiFi signals. In this framework, we transform segmentation tasks into classification problems and propose a CNN-based activity segmentation algorithm, which can reduce the dependence on experience and address the performance degradation problem. To further enhance the overall performance, we design a feedback mechanism, where the segmentation algorithm is refined based on the feedback computed using activity recognition results. The experiments demonstrate that DeepSeg acquires remarkable gains compared with state-of-the-art approaches.

As a cornerstone service for many Internet of Things applications, channel state information (CSI)-based activity recognition has received immense attention over recent years. However, recognition performance of general approaches might significantly decrease when applying the trained model to the left-out user whose CSI data are not used for model training. To overcome this challenge, we propose a semi-supervised generative adversarial network (GAN) for CSI-based activity recognition (CsiGAN). Based on the general semi-supervised GANs, we mainly design three components for CsiGAN to meet the scenarios that unlabeled data from left-out users are very limited and enhance recognition performance: 1) we introduce a new complement generator, which can use limited unlabeled data to produce diverse fake samples for training a robust discriminator; 2) for the discriminator, we change the number of probability outputs from $k+1$ into $2k+1$ (here, $k$ is the number of categories), which can help to obtain the correct decision boundary for each category; and 3) based on the introduced generator, we propose a manifold regularization, which can stabilize the learning process. The experiments suggest that CsiGAN attains significant gains compared to the state-of-the-art methods.

Smart homes expects to improve the convenience, comfort, and energy efficiency of the residents by connecting and controlling various appliances. As the personal information and computing hub for smart homes, smartphones allow people to monitor and control their homes anytime and anywhere. Therefore, the security and privacy of smartphones and the stored data are crucial in smart homes. To protect smartphones from potential attacks, various built-in sensors can be utilized for user authentication/identification and access control to achieve enhanced security. In this paper, we propose a framework, smartphone sensor user identification (SSUI), in order to facilitate user identification based on the relationships between different types of sensor data and smartphone users. Specifically in SSUI, the time and frequency features are extracted and learned separately using convolution neural network (CNN). The CNN outputs are then processed using recurrent neural network, according to several time bins. Using both of our own dataset (collected from 17 participants) and a publicly available dataset (i.e., Heterogeneity Dataset for Human Activity Recognition), we demonstrate the effectiveness of the proposed SSUI framework, where we achieve an accuracy rate of over 91.45% in various scenarios.

Device authentication is a critical and challenging issue for the emerging Internet of Things (IoT). One promising solution to authenticate IoT devices is to extract a fingerprint to perform device authentication by exploiting variations in the transmitted signal caused by hardware and manufacturing inconsistencies. In this paper, we propose a lightweight device authentication protocol [named speaker-to-microphone (S2M)] by leveraging the frequency response of a speaker and a microphone from two wireless IoT devices as the acoustic hardware fingerprint. S2M authenticates the legitimate user by matching the fingerprint extracted in the learning process and the verification process, respectively. To validate and evaluate the performance of S2M, we design and implement it in both mobile phones and PCs and the extensive experimental results show that S2M achieves both low false negative rate and low false positive rate in various scenarios under different attacks.