Fake or tampered images pose a real problem in today’s life. It is easy to unknowingly be drawn to an interesting image that is false. Recently, with the emergence of generative adversarial networks (GANs), it becomes much more easy to generate high-quality fake images in a very realistic way. However, the current digital image forensics algorithms mainly focus on the detection of traditional tampered images or need prior knowledge of the network structure of GANs. Hence, verifying the authenticity of an image is very challenging. In this paper, we propose a general method for simultaneously detecting tampered images, and GANs generated images. First, we use the Scharr operator to extract the edge information of the image. Then, we converted the edge image information matrix into the gray level co-occurrence matrix (GLCM) to scale the image without loss of image information. Finally, GLCM was fed into the deep neural network designed based on depthwise separable convolution for training. Compared with other methods, our model achieves a higher macro average of F1 score of 0.9865. Meanwhile, our method has better performance in detecting tampered images and has strong generalization ability for many GANs models.

With the rapid development of Internet of Things technology, cloud computing, and big data, the combination of medical systems and information technology has become increasingly close. However, the emergence of intelligent medical systems has brought a series of network security threats and hidden dangers, including data leakage and remote attacks, which can directly threaten patients’ lives. To ensure the security of medical information systems and expand the application of zero trust in the medical field, we combined the medical system with the zero-trust security system to propose a zero-trust medical security system. In addition, in its dynamic access control module, based on the RBAC model and the calculation of user behavior risk value and trust, an access control model based on subject behavior evaluation under zero-trust conditions (ABEAC) was designed to improve the security of medical equipment and data. Finally, the feasibility of the system is verified through a simulation experiment.

With the development of Internet technology, network security is under diverse threats. In particular, attackers can spread malicious uniform resource locators (URL) to carry out attacks such as phishing and spam. The research on malicious URL detection is significant for defending against these attacks. However, there are still some problems in the current research. For instance, malicious features cannot be extracted efficiently. Some existing detection methods are easy to evade by attackers. We design a malicious URL detection model based on a dynamic convolutional neural network (DCNN) to solve these problems. A new folding layer is added to the original multilayer convolution network. It replaces the pooling layer with the k-max-pooling layer. In the dynamic convolution algorithm, the width of feature mapping in the middle layer depends on the vector input dimension. Moreover, the pooling layer parameters are dynamically adjusted according to the length of the URL input and the depth of the current convolution layer, which is beneficial to extracting more in-depth features in a wider range. In this paper, we propose a new embedding method in which word embedding based on character embedding is leveraged to learn the vector representation of a URL. Meanwhile, we conduct two groups of comparative experiments. First, we conduct three contrast experiments, which adopt the same network structure and different embedding methods. The results prove that word embedding based on character embedding can achieve higher accuracy. We then conduct the other three experiences, which use the same embedding method proposed in this paper and use different network structures to determine which network is most suitable for our model. We verify that the model designed in this paper has the highest accuracy (98%) in detecting malicious URL through these experiences.

Recent years have witnessed a boom of connected medical devices, which brings security issues in the meantime. Medical imaging devices, an essential part of medical cyber-physical systems, play a vital role in modern hospitals and are often life-critical. However, security and privacy issues in these medical cyber-physical systems are sometimes ignored. We perform an empirical study on imaging devices to analyse the security of medical cyber-physical systems. To be precise, we design a threat model and propose prospective attack techniques for medical imaging devices. To tackle potential cyber threats, we introduce protection mechanisms, evaluate the effectiveness and efficiency of protection mechanisms as well as its interplay with attack techniques. To scoring security, we design a hierarchical system that provides actionable suggestions for imaging devices in different scenarios. We investigate 15 devices from 9 manufacturers to demonstrate empirical comprehension and real-world security issues.

Most text sentiment classification methods have the shortcomings of insufficient semantic understanding and low classification accuracy. To solve these problems, a sentiment classification model based on capsule network (SC-BiCapsNet) was proposed in this paper. In the stage of word vectorization, the G-Word model is presented to improve the ability of text information representation. In the stage of text semantic representation, the COL-Att model is presented to optimize the semantic coding structure and enhance the effect of text semantic expression. In the stage of text classification, the capsule network is devised, and its core dynamic routing algorithm can increase the weight value of important features, thus improving the accuracy of classification results. This paper evaluates the performance of SC-BiCapsNet model on two text datasets, IMDB and NLPCC 2014 dataset. The experimental results demonstrate that our method outperforms the results of other related models on IMDB dataset and gets the highest accuracy of 0.9238. Meanwhile, comparing with the best system in NLPCC2014-Task II, the F1-scsore of SC-BiCapsNet has increased by at least 1.5%, which proves the validity of SC-BiCapsNet. In the future work, we will try to reduce the time consumption and improve the generalization ability of the model.