In order to solve the failure of electricity anti-stealing detection device triggered by the noise mixed in high-frequency electricity stealing signals,a denoising method based on variational mode decomposition(VMD)an...In order to solve the failure of electricity anti-stealing detection device triggered by the noise mixed in high-frequency electricity stealing signals,a denoising method based on variational mode decomposition(VMD)and wavelet threshold denoising(WTD)was applied to extract the effective high-frequency electricity stealing signals.First,the signal polluted by noise was pre-decomposed using the VMD algorithm,the instantaneous frequency means of each pre-decomposed components was analyzed,so as to determine the optimal K value.The optimal K value was used to decompose the polluted signal into K intrinsic mode components,and the sensitive mode components were determined through the cross-correlation function.Next,each sensitive mode was reconstructed.Finally,the reconstructed signal denoised using the wavelet threshold to obtain the denoised signal.The simulation analysis and experimental results show that the proposed method is superior to the traditional VMD method,FFT method and EMD method,as it can effectively eliminate the noise and enhance the reliability of high-frequency electricity stealing signal detection.展开更多
In the year of 1768,the country was sent into a nation wide panic by a sorcery case named"soul stealing"in China,to which the emperor,officials and civilians at that time all responded differently.Based on t...In the year of 1768,the country was sent into a nation wide panic by a sorcery case named"soul stealing"in China,to which the emperor,officials and civilians at that time all responded differently.Based on the"soul stealing"case,this paper gives an analysis of the political environment and the bureaucratic system to explore the motives driven by power involved in the proceedings of the case,and in doing so the paper attempts to provide insights into the implications of the ancient power system that makes possible the abuse of authority.展开更多
在黑盒场景下,使用模型功能窃取方法生成盗版模型已经对云端模型的安全性和知识产权保护构成严重威胁。针对扰动和软化标签(变温)等现有的模型窃取防御技术可能导致模型输出中置信度最大值的类别发生改变,进而影响原始任务中模型性能的...在黑盒场景下,使用模型功能窃取方法生成盗版模型已经对云端模型的安全性和知识产权保护构成严重威胁。针对扰动和软化标签(变温)等现有的模型窃取防御技术可能导致模型输出中置信度最大值的类别发生改变,进而影响原始任务中模型性能的问题,提出一种基于暗知识保护的模型功能窃取防御方法,称为DKP(defending against model stealing attacks based on Dark Knowledge Protection)。首先,利用待保护的云端模型对测试样本进行处理,以获得样本的初始置信度分布向量;然后,在模型输出层之后添加暗知识保护层,通过分区变温调节softmax机制对初始置信度分布向量进行扰动处理;最后,得到经过防御的置信度分布向量,从而降低模型信息泄露的风险。使用所提方法在4个公开数据集上取得了显著的防御效果,尤其在博客数据集上使盗版模型的准确率降低了17.4个百分点,相比之下对后验概率进行噪声扰动的方法仅能降低约2个百分点。实验结果表明,所提方法解决了现有扰动、软化标签等主动防御方法存在的问题,在不影响测试样本分类结果的前提下,通过扰动云端模型输出的类别概率分布特征,成功降低了盗版模型的准确率,实现了对云端模型机密性的可靠保障。展开更多
Intrusion detection systems are increasingly using machine learning.While machine learning has shown excellent performance in identifying malicious traffic,it may increase the risk of privacy leakage.This paper focuse...Intrusion detection systems are increasingly using machine learning.While machine learning has shown excellent performance in identifying malicious traffic,it may increase the risk of privacy leakage.This paper focuses on imple-menting a model stealing attack on intrusion detection systems.Existing model stealing attacks are hard to imple-ment in practical network environments,as they either need private data of the victim dataset or frequent access to the victim model.In this paper,we propose a novel solution called Fast Model Stealing Attack(FMSA)to address the problem in the field of model stealing attacks.We also highlight the risks of using ML-NIDS in network security.First,meta-learning frameworks are introduced into the model stealing algorithm to clone the victim model in a black-box state.Then,the number of accesses to the target model is used as an optimization term,resulting in minimal queries to achieve model stealing.Finally,adversarial training is used to simulate the data distribution of the target model and achieve the recovery of privacy data.Through experiments on multiple public datasets,compared to existing state-of-the-art algorithms,FMSA reduces the number of accesses to the target model and improves the accuracy of the clone model on the test dataset to 88.9%and the similarity with the target model to 90.1%.We can demonstrate the successful execution of model stealing attacks on the ML-NIDS system even with protective measures in place to limit the number of anomalous queries.展开更多
BACKGROUND Portal vein arterialization(PVA)has been used in liver transplantation(LT)to maximize oxygen delivery when arterial circulation is compromised or has been used as an alternative reperfusion technique for co...BACKGROUND Portal vein arterialization(PVA)has been used in liver transplantation(LT)to maximize oxygen delivery when arterial circulation is compromised or has been used as an alternative reperfusion technique for complex portal vein thrombosis(PVT).The effect of PVA on portal perfusion and primary graft dysfunction(PGD)has not been assessed.All patients receiving PVA and LT at the Fundacion Santa Fe de Bogota between 2011 and 2022 were analyzed.To account for the time-sensitive effects of graft perfusion,patients were classified into two groups:prereperfusion(pre-PVA),if the arterioportal anastomosis was performed before graft revascularization,and postreperfusion(post-PVA),if PVA was performed afterward.The pre-PVA rationale contemplated poor portal hemodynamics,severe vascular steal,or PVT.Post-PVA was considered if graft hypoperfusion became evident.Conservative interventions were attempted before PVA.展开更多
基金supported by China Southern Power Grid Corporation,GrantNo.GDKJXM20185800the National Natural Science Foundation of China,Grant No.62073084.
文摘In order to solve the failure of electricity anti-stealing detection device triggered by the noise mixed in high-frequency electricity stealing signals,a denoising method based on variational mode decomposition(VMD)and wavelet threshold denoising(WTD)was applied to extract the effective high-frequency electricity stealing signals.First,the signal polluted by noise was pre-decomposed using the VMD algorithm,the instantaneous frequency means of each pre-decomposed components was analyzed,so as to determine the optimal K value.The optimal K value was used to decompose the polluted signal into K intrinsic mode components,and the sensitive mode components were determined through the cross-correlation function.Next,each sensitive mode was reconstructed.Finally,the reconstructed signal denoised using the wavelet threshold to obtain the denoised signal.The simulation analysis and experimental results show that the proposed method is superior to the traditional VMD method,FFT method and EMD method,as it can effectively eliminate the noise and enhance the reliability of high-frequency electricity stealing signal detection.
基金supported by the Key Discipline of Administrative Law Project of Shanghai University of Political Science and Law at Colleges and Universities sponsored by the Central Finance
文摘In the year of 1768,the country was sent into a nation wide panic by a sorcery case named"soul stealing"in China,to which the emperor,officials and civilians at that time all responded differently.Based on the"soul stealing"case,this paper gives an analysis of the political environment and the bureaucratic system to explore the motives driven by power involved in the proceedings of the case,and in doing so the paper attempts to provide insights into the implications of the ancient power system that makes possible the abuse of authority.
文摘在黑盒场景下,使用模型功能窃取方法生成盗版模型已经对云端模型的安全性和知识产权保护构成严重威胁。针对扰动和软化标签(变温)等现有的模型窃取防御技术可能导致模型输出中置信度最大值的类别发生改变,进而影响原始任务中模型性能的问题,提出一种基于暗知识保护的模型功能窃取防御方法,称为DKP(defending against model stealing attacks based on Dark Knowledge Protection)。首先,利用待保护的云端模型对测试样本进行处理,以获得样本的初始置信度分布向量;然后,在模型输出层之后添加暗知识保护层,通过分区变温调节softmax机制对初始置信度分布向量进行扰动处理;最后,得到经过防御的置信度分布向量,从而降低模型信息泄露的风险。使用所提方法在4个公开数据集上取得了显著的防御效果,尤其在博客数据集上使盗版模型的准确率降低了17.4个百分点,相比之下对后验概率进行噪声扰动的方法仅能降低约2个百分点。实验结果表明,所提方法解决了现有扰动、软化标签等主动防御方法存在的问题,在不影响测试样本分类结果的前提下,通过扰动云端模型输出的类别概率分布特征,成功降低了盗版模型的准确率,实现了对云端模型机密性的可靠保障。
基金supported by Grant Nos.U22A2036,HIT.OCEF.2021007,2020YFB1406902,2020B0101360001.
文摘Intrusion detection systems are increasingly using machine learning.While machine learning has shown excellent performance in identifying malicious traffic,it may increase the risk of privacy leakage.This paper focuses on imple-menting a model stealing attack on intrusion detection systems.Existing model stealing attacks are hard to imple-ment in practical network environments,as they either need private data of the victim dataset or frequent access to the victim model.In this paper,we propose a novel solution called Fast Model Stealing Attack(FMSA)to address the problem in the field of model stealing attacks.We also highlight the risks of using ML-NIDS in network security.First,meta-learning frameworks are introduced into the model stealing algorithm to clone the victim model in a black-box state.Then,the number of accesses to the target model is used as an optimization term,resulting in minimal queries to achieve model stealing.Finally,adversarial training is used to simulate the data distribution of the target model and achieve the recovery of privacy data.Through experiments on multiple public datasets,compared to existing state-of-the-art algorithms,FMSA reduces the number of accesses to the target model and improves the accuracy of the clone model on the test dataset to 88.9%and the similarity with the target model to 90.1%.We can demonstrate the successful execution of model stealing attacks on the ML-NIDS system even with protective measures in place to limit the number of anomalous queries.
文摘BACKGROUND Portal vein arterialization(PVA)has been used in liver transplantation(LT)to maximize oxygen delivery when arterial circulation is compromised or has been used as an alternative reperfusion technique for complex portal vein thrombosis(PVT).The effect of PVA on portal perfusion and primary graft dysfunction(PGD)has not been assessed.All patients receiving PVA and LT at the Fundacion Santa Fe de Bogota between 2011 and 2022 were analyzed.To account for the time-sensitive effects of graft perfusion,patients were classified into two groups:prereperfusion(pre-PVA),if the arterioportal anastomosis was performed before graft revascularization,and postreperfusion(post-PVA),if PVA was performed afterward.The pre-PVA rationale contemplated poor portal hemodynamics,severe vascular steal,or PVT.Post-PVA was considered if graft hypoperfusion became evident.Conservative interventions were attempted before PVA.
