Steganalysis of neural networks based on parameter statistical bias

Many pretrained deep learning models have been released to help engineers and researchers develop deep learning-based systems or conduct research with minimall effort. Previous work has shown that at secret message can be embedded in neural network parameters without compromising the accuracy of the model. Malicious developers can, therefore, hide malware or other baneful information in pretrained models, causing harm to society. Hence, reliable detection of these vicious pretrained models is urgently needed. We analyze existing approaches for hiding messages and find that they will ineluctably cause biases in the parameter statistics. Therefore, we propose steganalysis methods for steganography on neural network parameters that extract statistics from benign and malicious models and build classifiers based on the extracted statistics. To the best of our knowledge, this is the first study on neural network steganalysis. The experimental results reveal that our proposed algorithm can effectively detect a model with an embedded message. Notably, our detection methods are still valid in cases where the payload of the stego model is low.     

独蒜兰优质种苗快速培育研究

建立独蒜兰优质种苗快繁体系,为实现独蒜兰植物工业化育苗、保护野生植物资源和发展人工育种技术提供新的途径.本研究从不同基本培养基对独蒜兰种子萌发影响、丛生芽诱导增殖培养和无根苗生根壮苗等方面开展研究.筛选出影响独蒜兰种子萌发的最佳基本培养基为B5,培养30 d统计种子萌发率为92％;丛生芽诱导增殖的最佳培养基为B5+TDZ 2.0 mg/L+KT 0.5 mg/L+200.0 mg/L肌醇+2.0 g/L活性炭+20.0 g/L蔗糖,培养60 d丛生芽率诱导率为98.00％,平均产生丛生芽数为6.80个;最佳生根壮苗培养基为1/2MS+IBA 0.2 mg/L+6-BA 0.1 mg/L+2.0 g/L活性炭+30.0 g/L土豆泥+20.0 g/L蔗糖,培养60 d后统计生根率为100％,平均生根数为3.20条,平均鲜重为21.11 g.     

负面在线评论及商家回复对顾客购买意愿的影响

从负面在线评论比例和失误严重性两个角度出发，通过引入归因理论、感知信任和商家回复等相关内容，探究负面在线评论对潜在顾客购买意愿影响机制和内部机理。对496份调查问卷所得数据进行统计分析，研究结果表明：负面在线评论比例和失误严重性都对潜在顾客购买意愿有显著的负向影响，且两者都通过影响失误的归属性归因、可控性归因、稳定性归因和感知信任影响潜在顾客购买意愿；在负面在线评论比例与失误严重性不同的情况下，商家应采取不同的回复策略实现节约成本与减小损失的目的。     

孕早期睡眠时长与午睡频次对妊娠糖尿病的影响

目的探讨孕早期睡眠时长和午睡频次对妊娠糖尿病(GDM)的影响。方法采用病例对照研究设计,选取2018年4月至2019年5月在安徽省妇幼保健院、吉安市妇幼保健院产检的孕妇作为研究对象。通过问卷调查收集研究对象的睡眠和午睡等信息,GDM的诊断来自孕中期常规口服葡萄糖耐量试验结果。结果共纳入500名研究对象,其中病例组196人,对照组304人。多元Logistic回归分析结果显示,与睡眠时长7～8.9 h/晚相比,睡眠7、9～9.9、≥10 h/晚的孕妇发生GDM的风险均显著上升;与午睡频次≤5 d/周相比,5 d/周的孕妇发生GDM的风险更高。相比孕早期午睡频率≤5 d/周的孕妇,孕早期睡眠时长不足(7 h/晚)对GDM的影响在午睡频率5 d/周的孕妇中更弱,交互作用具有统计学意义(P=0.002)。结论孕早期睡眠时长、午睡频次均与GDM发病相关,且二者在GDM的发生中存在交互作用。     

基于区块链技术的生鲜农产品溯源研究

为探究区块链技术在生鲜农产品流通过程中如何保障质量安全,利用演化博弈方法探究监管部门与生鲜农产品生产商关于生鲜食品质量安全溯源博弈过程.构建监管部门与生鲜农产品生产商的质量安全溯源的博弈支付矩阵,并进一步确立质量安全溯源的复制动态方程.通过对二维复制动态系统进行求解初步得到系统的平衡点,再分析平衡点的局部稳定性情况.结果显示,当生产商生产质量安全生鲜产品的收益大于生产不安全生鲜产品的收益时,双方博弈将最终演化到理想状态(1,0),即生产商将生产质量安全的生鲜农产品而监管部门选择不监管.由此可知,当生产商生产质量不安全的产品会获得更多收益时,监管部门可以采取罚金的方式来规范生产商安全生产.     

肌无力方剂300剂处方分析

目的:调查我院门诊肌无力患者处方使用情况,了解治疗肌无力的用药现状并作出相应分析,为指导临床合理用药提供依据。方法:随机抽取我院知名专家况时祥教授2017年8月至2018年7月的门诊肌无力患者中药处方300张,对处方中的患者信息和用药情况进行统计分析。结果:肌无力患者主要集中在40～80岁之间,女性高发,处方药以补益类为主。其中,黄芪最为常用。结论:处方分析可为中医药治疗疾病管理模式提供理论支持,更有效指导临床合理用药。     

夯土遗址表层热劣化模拟试验研究

以武威明长城夯土遗址为例,通过现场调查及室内模拟试验,研究分析夯土遗址表层的热劣化机制.X射线衍射及电镜扫描试验表明:遗址表层与母墙土的矿物成分相似,表层微观结构平滑,但风化明显.热劣化模拟试验表明,试样表层热传导系数、热扩散系数及体积比热均低于母墙.热劣化过程中,表层与母墙的热传导系数与热扩散系数均维持稳定.研究认为,夯土遗址表层与母墙组成具有热性质差异的双层结构,在温差作用下,双层结构间的稳定热差异会使其接触面产生热应力差.热应力差的持续作用使接触面劣化疲劳,导致遗址表层与母墙分离,产生剥离病害.     

麻黄细辛附子汤6种活性成分提取动态变化研究

研究麻黄细辛附子汤中盐酸麻黄碱、盐酸伪麻黄碱、细辛脂素、苯甲酰新乌头原碱、苯甲酰乌头原碱、苯甲酰次乌头原碱在提取过程中的含量动态变化。采用HPLC，测定水提取过程中不同时间提取液中6种活性成分的含量。结果表明，6种活性成分溶出量均在前90 min呈上升趋势，且组分稳定存在。90 min后，溶出量增长趋势变缓并有不同程度的消减，存在状态不稳定。研究认为麻黄细辛附子汤的提取工艺为水煎提取90 min，既可保证6种活性成分的稳定存在，又使其溶出量达峰值。     

Real-Time Pricing of Smart Grid Based on Piece-Wise Linear Functions

In a power grid system, utility is a measure of the satisfaction of users' electricity consumption; cost is a monetary value of electricity generated by the supplier. The utility and cost functions represent the satisfaction of different users and the supplier. Quadratic utility, logarithmic utility,and quadratic cost functions are widely used in social welfare maximization models of real-time pricing. These functions are not universal; they have to be discussed in detail for individual models. To overcome this problem, a piece-wise linear utility function and a piece-wise linear cost function with general properties are proposed in this paper. By smoothing the piece-wise linear utility and cost functions, a social welfare maximization model can be transformed into a differentiable convex optimization problem. A dual optimization method is used to solve the smoothed model. Through mathematical deduction and numerical simulations, the rationality of the model and the validity of the algorithm are verified as long as the elastic and cost coefficients take appropriate values. Thus, different user types and the supplier can be determined by selecting different elastic and cost coefficients.