With the spread of Internet of Things (IoT) devices, countermeasures against cyber-attacks have become an issue. In this study, we focused on anomaly detection using flow data, which can reduce the data volume, and proposed a new anomaly detection method that combines a new graph composition method that represents a sequence of flow data as a graph and a graph neural network (GNN). Various detection methods, including deep learning, have been proposed for identifying malware such as denial-of-service (DoS) attacks, in which the characteristics of traffic deviate significantly from those of benign communications. We conducted an evaluation experiment with the proposed method using the KDDI-IoT-2019 dataset and discussed its effectiveness and limitations.

In this paper, we propose a simple and effective preprocessing method for melanoma classification by considering cytological properties of melanomas, in particular the alignment of the major axis of the tumor in the same direction. We evaluate our method with a set of 1,760 dermoscopic images (329 of melanomas and 1,431 of nevi) and a simple convolutional neural network (CNN) classifier with five-fold cross validation. The proposed tumor alignment method improves the classification performance by 5.8% in terms of the area under the ROC curve (AUC). In addition, it proves to be 2.1% better in term of AUC when compared with the same configured CNN trained using images that are nine times larger. Our results also show that considering the intrinsic features of the classification target is important even when the classifier has a capability to obtain effective features automatically through its learning process.

When smart phone users browse web news articles, they encounter attractive phrases by chance. At that time, they try to obtain information on the smart phones. In order to search the information of the attractive phrases on web pages, they have to manipulate the smart phones to search on the small screen. It causes stresses because of situations such as manipulation errors. Such stresses could be eliminated if the attractive phrases can be identified and be input automatically in order to recommend the web pages with the information of the attractive phrases. Because of the small screens of smart phones, users move the news article area displayed on the screens with touch gestures such as swipe. The history of touch gestures during browsing an article implies the position and the timing users have focused on in the article. This paper proposes a method to identify the areas where attractive phrases have appeared on news articles, in order to enable automatic identification of attractive phrases. We have achieved real-time identification by utilizing the history of touch gestures during a web news browsing. The proposed method shows the history of touch gestures by a gesture trail. It is a graph showing the time series of the vertical coordinate of the displayed article area. When users encounter attractive phrases, they take certain patterns of touch gestures to confirm or read carefully the neighborhood of the attractive phrases. Thus, in the proposed method, the news article area including an attractive phrase is detected by matching the gesture trail in a sliding time window with a pattern obtained by pre-training. We use slow-down and resting patterns approximated by quadratic functions with the parameters defined for individual users. Experiments to identify time windows of attractive phrases on gesture trails has revealed that the highest and the lowest precision ratios are 0.579 and 0.278, respectively.

This article shows that /spl epsiv/-almost strongly universal classes of hash functions can yield better explicit constructions of identification codes via channels (ID codes) and identification plus transmission codes (IT codes) than the previous explicit constructions of Verdu and Wei (1993).