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.     

Banach 3-Lie代数上两类映射的稳定性

结合函数方程pf(x+y/p+z)=f(x)+f(y)+pf(z),研究了Banach 3-Lie代数上的同态、导子以及广义导子的广义Hyers-Ulam稳定性(其中p为固定正整数).     

基于知识图谱的核电设备健康管理知识建模与分析

随着信息技术发展,设备健康数据与知识图谱技术结合为设备健康发展带来新的发展机遇,知识图谱利用其多种特异性优势,使企业系统中积累的海量设备维修数据得到有效整合利用.本文提出基于知识图谱的核电设备健康管理知识建模与分析方法,并根据实体关系模型构建维修知识图谱本体框架,同时基于核电企业实际的维修工单数据,从统计分析和关联分析两方面取得良好的实践效果.研究结果表明,知识图谱能够在设备维修知识的集成上解决传统的数据孤岛问题,同时能基于故障的可视化分析为维修人员和管理人员提供决策和知识支持.     

自动驾驶汽车车底边框的优化算法

为了对通过YOLO框架所提取出的汽车底边框进行优化,首先对于道路上的每一辆汽车底部部分像素块建立改进后的高斯混合模型,通过该模型来刻画道路灰度值以及车底阴影灰度值的正态分布规律,然后通过对每辆检测到的汽车进行自上而下的扫描算法以确定汽车边框的实际高度,最后通过进行路测实验来说明算法的可行性与准确性.     

Impact of the RMB Joining in the SDR Basket on Its Internationalization from the Perspective of Risk Spillover

For evaluating the influence of the Chinese renminbi(RMB) joining in the special drawing right(SDR) basket on RMB's internationalization, the authors systemically study the risk spillover networks and examine the dynamic relationship of exchange rates among the SDR currencies including the US dollar(USD), European Union euro(EUR), Japanese yen(JPY) and British pound(GBP).The empirical results demonstrate that the USD takes a dominant position and holds the biggest risk spillover to other currencies, and the RMB's inclusion to the SDR basket makes the risk spillover to get average, giving rise to the SDR currency system more stable to a certain degree. The inclusion of the RMB in the SDR not only can reduce the systematic risk of the SDR, but also has a certain impact on the international exchange rate markets. Nowadays, in front of the growing trade friction, more such researches could help to effectively deal with the currency disputes.     

双向认知网络物理层安全策略建模及仿真研究

为提高移动通信系统的性能,构建了一个存在主网络干扰和窃听者的双向认知中继网络的物理层安全传输模型,设计了一种中继选择和功率分配的联合优化策略,用于对抗窃听者,保护收发节点的信息传输。通过仿真分析得出,此非凸优化问题限制条件较多,难以保证粒子群优化算法随机生成的初始解的可行性,从而造成求解困难。提出了一种基于可变网格优化和粒子群优化算法的混合优化算法。仿真结果表明此算法提高了次级网络的保密速率,提升了次级网络的安全性能。     

南京长江第四大桥钢箱梁铺装层温度影响系数分析

为了减少现有的钢桥面铺装层车辙深度预估的误差,提出基于长期实测数据的钢桥面铺装车辙评估模型的温度影响系数计算公式。首先对南京长江第四大桥的实测温度概率统计后采用双高斯函数表征其分布;在此基础上修正了现有车辙评估模型中的温度影响系数公式,并利用车辙实测数据对车辙评估模型计算值进行验证。研究结果表明:钢箱梁复合浇筑式铺装层温度可通过双高斯分布描述其概率分布;计入温度影响系数的车辙深度与实际车辙值之间误差较小,相对误差为15.2%。     

大尺寸工程模型试验中的相似材料配比试验研究

大型物理模型试验是研究复杂工程问题的重要方法，如何快速、准确地确定相似材料配比是试验中至关重要的一环.为降低试验成本、简化试验步骤、充分调用原料性能，采用河砂、水泥和石膏这三种最普通的原料，以骨胶比(河砂与水泥石膏的质量比)和水膏比(水泥与石膏的质量比)为变量，进行了45组对比试验，对相似材料的物理及动、静力学参数进行了研究.获得了以河砂、水泥和石膏为原料配制的相似材料各物理量的变化区间；探讨了骨胶比和水膏比对相似材料各物理量的控制作用，并对比了其敏感性；基于多元线性回归分析建立了各因素与物理量间的定量关系，得到的拟合公式能较好地描述相似材料的性质.     

Blockage of the NLRP3 inflammasome by MCC950 improves anti-tumor immune responses in head and neck squamous cell carcinoma

The NLRP3 inflammasome is a critical innate immune pathway responsible for producing active interleukin (IL)-1β, which is associated with tumor development and immunity. However, the mechanisms regulating the inflammatory microenvironment, tumorigenesis and tumor immunity are unclear. Herein, we show that the NLRP3 inflammasome was over-expressed in human HNSCC tissues and that the IL-1β concentration was increased in the peripheral blood of HNSCC patients. Additionally, elevated NLRP3 inflammasome levels were detected in tumor tissues of Tgfbr1/Pten 2cKO HNSCC mice, and elevated IL-1β levels were detected in the peripheral blood serum, spleen, draining lymph nodes and tumor tissues. Blocking NLRP3 inflammasome activation using MCC950 remarkably reduced IL-1β production in an HNSCC mouse model and reduced the numbers of myeloid-derived suppressor cells (MDSCs), regulatory T cells (Tregs) and tumor-associated macrophages (TAMs). Moreover, inhibiting NLRP3 inflammasome activation increased the numbers of CD4⁺ and CD8⁺ T cells in HNSCC mice. Notably, the numbers of exhausted PD-1⁺ and Tim3⁺ T cells were significantly reduced. A human HNSCC tissue microarray showed that NLRP3 inflammasome expression was correlated with the expression of CD8 and CD4, the Treg marker Foxp3, the MDSC markers CD11b and CD33, and the TAM markers CD68 and CD163, PD-1 and Tim3. Overall, our results demonstrate that the NLRP3 inflammasome/IL-1β pathway promotes tumorigenesis in HNSCC and inactivation of this pathway delays tumor growth, accompanied by decreased immunosuppressive cell accumulation and an increased number of effector T cells. Thus, inhibition of the tumor microenvironment through the NLRP3 inflammasome/IL-1β pathway may provide a novel approach for HNSCC therapy.