An Ensemble Detection Method for Shilling Attacks Based on Features of Automatic Extraction

Faced with the evolving attacks in recommender systems, many detection features have been proposed by human engineering and used in supervised or unsupervised detection methods. However, the detection features extracted by human engineering are usually aimed at some specific types of attacks. To further detect other new types of attacks, the traditional methods have to re-extract detection features with high knowledge cost. To address these limitations, the method for automatic extraction of robust features is proposed and then an Adaboost-based detection method is presented. Firstly, to obtain robust representation with prior knowledge, unlike uniform corruption rate in traditional mLDA(marginalized Linear Denoising Autoencoder), different corruption rates for items are calculated according to the ratings’ distribution. Secondly, the ratings sparsity is used to weight the mapping matrix to extract low-dimensional representation. Moreover, the uniform corruption rate is also set to the next layer in mSLDA(marginalized Stacked Linear Denoising Autoencoder) to extract the stable and robust user features. Finally, under the robust feature space, an Adaboost-based detection method is proposed to alleviate the imbalanced classification problem. Experimental results on the Netflix and Amazon review datasets indicate that the proposed method can effectively detect various attacks.

Disturbance identification of electric field data observed by the CSES-01 satellite before earthquakes

Numerous studies have confirmed that electromagnetic disturbances before earthquakes can be observed by satellites.In this study,we use the C-value method that includes the acoustic whistle signature;pre-seismic ionospheric electromagnetic disturbance signals were acquired based on the CSES-01 satellite electric field data,and the maximum value of C in the earthquake preparation zones increased continuously from 2.0 three days before the earthquake and reached a maximum weight of 3.0 on the day of the earthquake,after the earthquake,it gradually decreased and recovered to about 2.0;its the C values fluctuated between-2 and 3,it is different from the C values range-2–12 of the previous seismic case study using the DEMETER satellite,which may be related to the orbital altitude and revisit period of the satellite.Then,the C values were normalized,and the time series analysis of the obtained θ values were done,and the results showed that:In the pregnant zone,the background variation of the disturbance amplitude θ is within 2σ,and the maximum disturbance amplitude of θ starts to increase gradually from the seventh period(one period of 5 days,i.e.,35–39 days before the earthquake),it reached 2σ by the fourth preseismic cycle(20–24 days before the earthquake),and then dropped sharply to about 1.5σ in the third pre-seismic cycle(15–19days before the earthquake),after two cycles of increase,the θ over the epicenter reached a maximum of 2.1σ at the time of the earthquake(combining the time of the earthquake and the satellite flight characteristics,the epicenter period is defined as January25-January 29,2020,and this defines the study time period line),and the θ decreases to within 2 times the standard range after the earthquake;The negative value of the disturbance amplitude θ in the central region of the pregnant seismic zone during the earthquake shows the transient energy release process.Through comparison,the θ values obtained by normalization based on the C-value method takes into account the variation of the background field,and the result can better reflect the energy change of the ionospheric field before the earthquakes.