The Internet of Things has broad application in military field, commerce, environmental monitoring, and many other fields. However, the open nature of the information media and the poor deployment environment have brought great risks to the security of wireless sensor networks, seriously restricting the application of wireless sensor network. Internet of Things composed of wireless sensor network faces security threats mainly from Dos attack, replay attack, integrity attack, false routing information attack, and flooding attack. In this paper, we proposed a new intrusion detection system based onK-nearest neighbor (K-nearest neighbor, referred to as KNN below) classification algorithm in wireless sensor network. This system can separate abnormal nodes from normal nodes by observing their abnormal behaviors, and we analyse parameter selection and error rate of the intrusion detection system. The paper elaborates on the design and implementation of the detection system. This system has achieved efficient, rapid intrusion detection by improving the wireless ad hoc on-demand distance vector routing protocol (Ad hoc On-Demand Distance the Vector Routing, AODV). Finally, the test results show that: the system has high detection accuracy and speed, in accordance with the requirement of wireless sensor network intrusion detection.

Web service is one of the key communications software services for the Internet. Web phishing is one of many security threats to web services on the Internet. Web phishing aims to steal private information, such as usernames, passwords, and credit card details, by way of impersonating a legitimate entity. It will lead to information disclosure and property damage. This paper mainly focuses on applying a deep learning framework to detect phishing websites. This paper first designs two types of features for web phishing: original features and interaction features. A detection model based on Deep Belief Networks (DBN) is then presented. The test using real IP flows from ISP (Internet Service Provider) shows that the detecting model based on DBN can achieve an approximately 90% true positive rate and 0.6% false positive rate.

The password has become today's dominant method of authentication. While brute-force attack methods such as HashCat and John the Ripper have proven unpractical, the research then switches to password guessing. State-of-the-art approaches such as the Markov Model and probabilistic context-free grammar (PCFG) are all based on statistical probability. These approaches require a large amount of calculation, which is time-consuming. Neural networks have proven more accurate and practical in password guessing than traditional methods. However, a raw neural network model is not qualified for cross-site attacks because each dataset has its own features. Our work aims to generalize those leaked passwords and improves the performance in cross-site attacks.

In recent years, more and more anonymous network have been developed. Since user's identity is difficult to trace in anonymous networks, many illegal activities are carried out in darknet. In this paper, we propose a hierarchical classifier of darknet traffic which can distinguish four types of darknet(Tor, I2P, ZeroNet, Freenet) and 25 darknet users' behavior. Due to the lack of public datasets, we deployed a darknet data probe that can capture real darknet traffic in Tor, I2P, ZeroNet, Freenet. After collecting and labeling darknet traffic, we extract 26 time-based flow features that can represent the characteristics of darknet traffic and train a hierarchical classifier constructed by 6 local classifiers. Results show that the classifier can easily distinguish Tor, I2P, ZeroNet, Freenet four kinds of darknet clients with an accuracy of 96.9% and identify 8 kinds of user behaviors for each type of darknet with an accuracy of 91.6% on average. With the help of this hierarchical classification method, darknet user behaviors can be accurately distinguished at the traffic exit.

A deep learning computer artificial intelligence system is helpful for early identification of ground glass opacities (GGOs). Images from the Lung Image Database Consortium and Image Database Resource Initiative (LIDC-IDRI) database were used in AlexNet and GoogLeNet to detect pulmonary nodules, and 221 GGO images provided by Xinhua Hospital were used in ResNet50 for detecting GGOs. We used computed tomography image radial reorganization to create the input image of the three-dimensional features, and used the extracted features for deep learning, network training, testing, and analysis. In the final evaluation results, we found that the accuracy of identification of lung nodule could reach 88.0%, with an F-score of 0.891. In terms of performance and accuracy, our method was better than the existing solutions. The GGO nodule classification achieved the best F-score of 0.87805. We propose a preprocessing method of red, green, and blue (RGB) superposition in the region of interest to effectively increase the differentiation between nodules and normal tissues, and that is the innovation of our research. The method of deep learning proposed in this study is more sensitive than other systems in recent years, and the average false positive is lower than that of others.

Malware remains a major threat to nowadays Internet. In this paper, we propose a DNS graph mining-based malware detection approach. A DNS graph is composed of DNS nodes, which represent server IPs, client IPs, and queried domain names in the process of DNS resolution. After the graph construction, we next transform the problem of malware detection to the graph mining task of inferring graph nodes’ reputation scores using the belief propagation algorithm. The nodes with lower reputation scores are inferred as those infected by malwares with higher probability. For demonstration, we evaluate the proposed malware detection approach with real-world dataset. Our real-world dataset is collected from campus DNS servers for three months and we built a DNS graph consisting of 19,340,820 vertices and 24,277,564 edges. On the graph, we achieve a true positive rate 80.63% with a false positive rate 0.023%. With a false positive of 1.20%, the true positive rate was improved to 95.66%. We detected 88,592 hosts infected by malware or C&C servers, accounting for the percentage of 5.47% among all hosts. Meanwhile, 117,971 domains are considered to be related to malicious activities, accounting for 1.5% among all domains. The results indicate that our method is efficient and effective in detecting malwares.