Reversible solid oxide cell(RSOC)is a new energy conversion device with significant applications,especially for power grid peaking shaving.However,the reversible conversion process of power generation/energy storage p...Reversible solid oxide cell(RSOC)is a new energy conversion device with significant applications,especially for power grid peaking shaving.However,the reversible conversion process of power generation/energy storage poses challenges for the performance and stability of air electrodes.In this work,a novel high-entropy perovskite oxide La_(0.2)Pr_(0.2)Gd_(0.2)Sm_(0.2)Sr_(0.2)Co_(0.8)Fe_(0.2)O_(3−δ)(HE-LSCF)is proposed and investigated as an air electrode in RSOC.The electrochemical behavior of HE-LSCF was studied as an air electrode in both fuel cell and electrolysis modes.The polarization impedance(Rp)of the HE-LSCF electrode is only 0.25Ω·cm^(2) at 800℃ in an air atmosphere.Notably,at an electrolytic voltage of 2 V and a temperature of 800℃,the current density reaches up to 1.68 A/cm^(2).The HE-LSCF air electrode exhibited excellent reversibility and stability,and its electrochemical performance remains stable after 100 h of reversible operation.With these advantages,HE-LSCF is shown to be an excellent air electrode for RSOC.展开更多
针对现有无数据模型窃取攻击技术在有限的查询预算下难以拟合原始训练集分布,进而影响对目标模型决策边界拟合效果问题,提出了一种基于Stable Diffusion的模型窃取攻击方法(Model Extraction Attack Based on Stable Diffusion,MEASD)....针对现有无数据模型窃取攻击技术在有限的查询预算下难以拟合原始训练集分布,进而影响对目标模型决策边界拟合效果问题,提出了一种基于Stable Diffusion的模型窃取攻击方法(Model Extraction Attack Based on Stable Diffusion,MEASD).利用预训练的Stable Diffusion生成训练数据可能涵盖多个域,并包含大量非判别性样本,设计了ILAF方法以优化Stable Diffusion生成的数据品质.将高质量合成数据的原始样本与由对抗样本生成器生成的对抗样本组成替代训练集.由DPA模块组合的替代模型基于替代训练集拟合目标模型的决策边界.实验结果表明,在四个主流的基准数据集上与EBFA和DMEAE方法相比,所提的MEASD方法能够将目标模型决策边界的拟合程度提高至84%,对目标模型的黑盒对抗攻击成功率超过68%,并且查询预算较低.MEASD方法能够有效地提升目标模型决策边界拟合效果及攻击成功率.展开更多
基金supported by Fundamental Research Funds for the Central Universities(2023KYJD1008)the Science Research Projects of the Anhui Higher Education Institutions of China(2022AH051582).
文摘Reversible solid oxide cell(RSOC)is a new energy conversion device with significant applications,especially for power grid peaking shaving.However,the reversible conversion process of power generation/energy storage poses challenges for the performance and stability of air electrodes.In this work,a novel high-entropy perovskite oxide La_(0.2)Pr_(0.2)Gd_(0.2)Sm_(0.2)Sr_(0.2)Co_(0.8)Fe_(0.2)O_(3−δ)(HE-LSCF)is proposed and investigated as an air electrode in RSOC.The electrochemical behavior of HE-LSCF was studied as an air electrode in both fuel cell and electrolysis modes.The polarization impedance(Rp)of the HE-LSCF electrode is only 0.25Ω·cm^(2) at 800℃ in an air atmosphere.Notably,at an electrolytic voltage of 2 V and a temperature of 800℃,the current density reaches up to 1.68 A/cm^(2).The HE-LSCF air electrode exhibited excellent reversibility and stability,and its electrochemical performance remains stable after 100 h of reversible operation.With these advantages,HE-LSCF is shown to be an excellent air electrode for RSOC.
文摘针对现有无数据模型窃取攻击技术在有限的查询预算下难以拟合原始训练集分布,进而影响对目标模型决策边界拟合效果问题,提出了一种基于Stable Diffusion的模型窃取攻击方法(Model Extraction Attack Based on Stable Diffusion,MEASD).利用预训练的Stable Diffusion生成训练数据可能涵盖多个域,并包含大量非判别性样本,设计了ILAF方法以优化Stable Diffusion生成的数据品质.将高质量合成数据的原始样本与由对抗样本生成器生成的对抗样本组成替代训练集.由DPA模块组合的替代模型基于替代训练集拟合目标模型的决策边界.实验结果表明,在四个主流的基准数据集上与EBFA和DMEAE方法相比,所提的MEASD方法能够将目标模型决策边界的拟合程度提高至84%,对目标模型的黑盒对抗攻击成功率超过68%,并且查询预算较低.MEASD方法能够有效地提升目标模型决策边界拟合效果及攻击成功率.
