Deep Residual Networks have recently been shown to significantly improve the performance of neural networks trained on ImageNet, with results beating all previous methods on this dataset by large margins in the image classification task. However, the meaning of these impressive numbers and their implications for future research are not fully understood yet. In this survey, we will try to explain what Deep Residual Networks are, how they achieve their excellent results, and why their successful implementation in practice represents a significant advance over existing techniques. We also discuss some open questions related to residual learning as well as possible applications of Deep Residual Networks beyond ImageNet. Finally, we discuss some issues that still need to be resolved before deep residual learning can be applied on more complex problems.

This paper provides an extensive and complete survey on the process of detecting and preventing various types of IoT‐based security attacks. It is designed for software developers, researchers, and practitioners in the Internet of Things field who aim to understand the process of detecting and preventing these attacks. For each entry identified from the list, a brief description is provided along with references where more information can be found. However, We surveyed the current state‐of‐the‐art IoT security solutions and focused on four main aspects: (1) handpicking representative attacks, (2) identifying potential solutions, (3) performing a threat analysis for each attack and solution, and (4) ranking solutions according to the threats they overcome. By adopting this framework, we identified five main categories of defense mechanisms: distributed denial of service detection/prevention, default password protection, encryption mechanisms, intrusion detection/prevention, and anomaly detection. These solutions are relatively mature in terms of utility and usability. However, the security analysis is conducted only concerning specific attacks, which may or may not be relevant to real‐world deployment. Appropriate IoT security solutions should incorporate threat modeling while considering other factors such as resource consumption and implementation effort. Overall, evaluation of IoT security solutions is arduous due to the complexity of IoT OSes, heterogeneous IoT devices (e.g., various hardware platforms), limited availability of open‐source codebases, and restrictive policies towards intellectual property disclosure. In addition, we note that there remains a lack of studies that perform a systematic evaluation of the state‐of‐the‐art in terms of both frameworks/methodologies and mechanisms proposed.

Different from ship detection from synthetic aperture radar (SDSAR) and ship detection from spaceborne optical images (SDSOI), ship detection from visual image (SDVI) has better detection accuracy and real-time performance, which can be widely used in port management, cross-border ship detection, autonomous ship, safe navigation, and other real-time applications. In this paper, we proposed a new SDVI algorithm, named enhanced YOLO v3 tiny network for real-time ship detection. The algorithm can be used in video surveillance to realize the accurate classification and positioning of six types of ships (including ore carrier, bulk cargo carrier, general cargo ship, container ship, fishing boat, and passenger ship) in real-time. Based on the original YOLO v3 tiny network, we have made the following fine tunings. 1) The preset anchors trained on Seaship annotation data have the similar “dumpy” shape as the normal ships, helping the network to achieve faster and better training; 2) Convolution layer instead of max-pooling layer and expanding the channels of prediction network improve the small target detection ability of the algorithm. 3) Due to the problem that large-scale ships are easily disturbed by the onshore building, complex waves and light on the water surface, we introduced attention module named CBAM into the backbone network, which make the model more focused on the target. The detection accuracy of the proposed algorism is obviously better than that of the original YOLO v3 tiny work. Although it is slightly inferior to the Yolo v3 network, it has faster speed than Yolo v3. However, the proposed algorithm is a better trade-off between real-time performance and detection accuracy, and is more suitable for actual scenes. Compared with the SOAT algorithm in Z. Shao et al. (2020), our algorithm has a 9.6% improvement in mAP and a faster speed.

With the fast development of the Internet of Things (IoT) technologies, more IoT devices are currently connected with the Internet, resulting in more exchange of information. However, data privacy and security threats have become emerging challenges of the IoT. In this paper, we are concerned about the security of image transmission in green IoT. Image encryption algorithms for green IoT are faced with two challenges: 1) To save the cost, devices in green Internet of Things (IoT) always have very low computing power, so they cannot support high-precision computing and 2) The algorithm deployed on the device should be efficient to guarantee the device to run for a long time. To solve the problem of precision limited, we propose a parallel chaotic system named PSBP. ¹ The PSBP is composed of Piecewise Linear Chaotic Map (PWLCM), Skew Tent Map (STM) and Bernoulli map in parallel model with the limited precision of 16-bit, which can generate good-performance key matrix for encryption. Then, we introduce a cost-efficient image encryption algorithm based on PSBP (named IEPSBP) which addresses the second challenge. Instead of bits or bytes, we use one row or one column as the basic unit to realize permutation and diffusion. The security analysis and performance experiments indicate that IEPSBP is secure and cost-efficient for green IoT.

Deep neural networks (DNNs) have been widely adopted but they are vulnerable to intentionally crafted adversarial examples. Various attack methods against DNNs have been proposed, yet there still lacks theoretical explanation of adversarial examples. In this paper, we aim to understand adversarial examples from the attacking process and we assume adding perturbations to the key/sensitive regions of the image could fool image classification DNNs. We propose gradient shielding to verify the assumption which ignores insensitive information during generating adversarial examples. Specifically, we propose interactive gradient shielding (IGS) method which selects sensitive regions and then applies gradient-based attack. To remove region selection, we propose adaptive gradient shielding (AGS) method which ignores insensitive gradients automatically. We conduct extensive experiments to evaluate the performance and the results also corroborate our perspective. With this method, we won the first place in IJCAI-AAAC 2019 Non-targeted Adversarial Attack competition.

With the rapid development of information technologies, the security of cyberspace has become increasingly serious. Network intrusion detection is a practical scheme to protect network systems from cyber attacks. However, as new vulnerabilities and unknown attack types are constantly emerging, only a few samples of such attacks can be captured for analysis, which cannot be handled by the existing detection methods deployed in real systems. To handle this problem, we propose a few-shot class-incremental learning method called Branch Fusion Strategy based Network Intrusion Detection (BFS-NID for short), which can continuously learn new attack classes with only a few samples. BFS-NID includes a feature extractor module and a branch classifier learning module. The feature extractor module uses a vision transformer to learn better feature representations in a self-supervised manner, and the parameters of the feature extractor are fixed to avoid catastrophic forgetting when the model learns incrementally. The branch classifier learning module sets re-projection for different branch sessions to enhance the feature representation ability between classes and employs a branch fusion strategy to associate the context of learned attack classes with new classes in different sessions. We conducted extensive experiments on two popular network intrusion detection benchmark datasets (CIC-IDS2017 and CSE-CIC-IDS2018) and the results demonstrate that BFS-NID surpasses the baselines and achieves the best performance.