The defense techniques for machine learning are critical yet challenging due tothe number and type of attacks for widely applied machine learning algorithms aresignificantly increasing. Among these attacks, the poison...The defense techniques for machine learning are critical yet challenging due tothe number and type of attacks for widely applied machine learning algorithms aresignificantly increasing. Among these attacks, the poisoning attack, which disturbsmachine learning algorithms by injecting poisoning samples, is an attack with the greatestthreat. In this paper, we focus on analyzing the characteristics of positioning samples andpropose a novel sample evaluation method to defend against the poisoning attack cateringfor the characteristics of poisoning samples. To capture the intrinsic data characteristicsfrom heterogeneous aspects, we first evaluate training data by multiple criteria, each ofwhich is reformulated from a spectral clustering. Then, we integrate the multipleevaluation scores generated by the multiple criteria through the proposed multiplespectral clustering aggregation (MSCA) method. Finally, we use the unified score as theindicator of poisoning attack samples. Experimental results on intrusion detection datasets show that MSCA significantly outperforms the K-means outlier detection in terms ofdata legality evaluation and poisoning attack detection.展开更多
In the evaluation of some simulation systems, only small samples data are gotten due to the limited conditions. In allusion to the evaluation problem of small sample data, an interval estimation approach with the impr...In the evaluation of some simulation systems, only small samples data are gotten due to the limited conditions. In allusion to the evaluation problem of small sample data, an interval estimation approach with the improved grey confidence degree is proposed.On the basis of the definition of grey distance, three kinds of definition of the grey weight for every sample element in grey estimated value are put forward, and then the improved grey confidence degree is designed. In accordance with the new concept, the grey interval estimation for small sample data is deduced. Furthermore,the bootstrap method is applied for more accurate grey confidence interval. Through resampling of the bootstrap, numerous small samples with the corresponding confidence intervals can be obtained. Then the final confidence interval is calculated from the union of these grey confidence intervals. In the end, the simulation system evaluation using the proposed method is conducted. The simulation results show that the reasonable confidence interval is acquired, which demonstrates the feasibility and effectiveness of the proposed method.展开更多
Leadership in Energy and Environmental Design Commercial Interiors (LEED-CI) is more relevant to interior design, which, according to the sharing layer concept, differs from exterior design (which is usually evaluated...Leadership in Energy and Environmental Design Commercial Interiors (LEED-CI) is more relevant to interior design, which, according to the sharing layer concept, differs from exterior design (which is usually evaluated with the LEED New Construction sub-scheme). LEED-CI requires separate empirical analyses of LEED-CI certified buildings to further improve this sub-scheme. Therefore, in this study, Silver and Gold projects certified under LEED-CI-2009 in 14 US states were considered. Three project performance analyses, (i) certification, (ii) category, and (iii) cross-certification, were studied. The following results were revealed: (i) the range of the medians for Silver- and Gold-certified projects were 51-57 pts and 62-71 pts, respectively;(ii) in both Silver- and Gold-certified projects, Sustainable Sites (SS), Water Efficiency (WE), and Innovation in Design (ID) were the best-performing;Energy and Atmosphere (EA) and Indoor Environmental Quality (EQ) were intermediate-performing;and Material and Resources (MR) was the worst-performing categories;and (iii) in Silver-Gold cross-certification, category-focused (in 10 of 14 states) and category-unfocused (in four of 14 states) strategies were determined;in the category-focused strategy, the highest popular category was EA;the intermediate popular categories were WE, MR, and ID;and the lowest popular category was SS. Pooling all projects and all states into one frame can lead to the obscurement of the actual LEED-CI-2009 strategy(ies) in the transition from Silver to Gold certification in the US.展开更多
文摘The defense techniques for machine learning are critical yet challenging due tothe number and type of attacks for widely applied machine learning algorithms aresignificantly increasing. Among these attacks, the poisoning attack, which disturbsmachine learning algorithms by injecting poisoning samples, is an attack with the greatestthreat. In this paper, we focus on analyzing the characteristics of positioning samples andpropose a novel sample evaluation method to defend against the poisoning attack cateringfor the characteristics of poisoning samples. To capture the intrinsic data characteristicsfrom heterogeneous aspects, we first evaluate training data by multiple criteria, each ofwhich is reformulated from a spectral clustering. Then, we integrate the multipleevaluation scores generated by the multiple criteria through the proposed multiplespectral clustering aggregation (MSCA) method. Finally, we use the unified score as theindicator of poisoning attack samples. Experimental results on intrusion detection datasets show that MSCA significantly outperforms the K-means outlier detection in terms ofdata legality evaluation and poisoning attack detection.
文摘In the evaluation of some simulation systems, only small samples data are gotten due to the limited conditions. In allusion to the evaluation problem of small sample data, an interval estimation approach with the improved grey confidence degree is proposed.On the basis of the definition of grey distance, three kinds of definition of the grey weight for every sample element in grey estimated value are put forward, and then the improved grey confidence degree is designed. In accordance with the new concept, the grey interval estimation for small sample data is deduced. Furthermore,the bootstrap method is applied for more accurate grey confidence interval. Through resampling of the bootstrap, numerous small samples with the corresponding confidence intervals can be obtained. Then the final confidence interval is calculated from the union of these grey confidence intervals. In the end, the simulation system evaluation using the proposed method is conducted. The simulation results show that the reasonable confidence interval is acquired, which demonstrates the feasibility and effectiveness of the proposed method.
文摘Leadership in Energy and Environmental Design Commercial Interiors (LEED-CI) is more relevant to interior design, which, according to the sharing layer concept, differs from exterior design (which is usually evaluated with the LEED New Construction sub-scheme). LEED-CI requires separate empirical analyses of LEED-CI certified buildings to further improve this sub-scheme. Therefore, in this study, Silver and Gold projects certified under LEED-CI-2009 in 14 US states were considered. Three project performance analyses, (i) certification, (ii) category, and (iii) cross-certification, were studied. The following results were revealed: (i) the range of the medians for Silver- and Gold-certified projects were 51-57 pts and 62-71 pts, respectively;(ii) in both Silver- and Gold-certified projects, Sustainable Sites (SS), Water Efficiency (WE), and Innovation in Design (ID) were the best-performing;Energy and Atmosphere (EA) and Indoor Environmental Quality (EQ) were intermediate-performing;and Material and Resources (MR) was the worst-performing categories;and (iii) in Silver-Gold cross-certification, category-focused (in 10 of 14 states) and category-unfocused (in four of 14 states) strategies were determined;in the category-focused strategy, the highest popular category was EA;the intermediate popular categories were WE, MR, and ID;and the lowest popular category was SS. Pooling all projects and all states into one frame can lead to the obscurement of the actual LEED-CI-2009 strategy(ies) in the transition from Silver to Gold certification in the US.
