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.     

基于无人机实时数据多波次任务规划模型分析

针对导弹部队多波次作战任务规划问题, 依据无人机的实时数据, 构建了基于路径的多层规划模型, 并设计了模型的算法求解流程。使用遗传算法与禁忌搜索混合算法, 得出了任务规划中的最优路径规划, 并在此基础上进行了冲突的消除。通过仿真案例表明, 用无人机协同配合导弹部队作战, 实时传输作战数据, 能够解决战场信息模糊不确定的问题; 使用多层规划模型能够为导弹多波次规划作战的路径进行科学的决策和选择。利用遗传算法和禁忌搜索混合算法, 能够避免局部最优导致无法输出结果的现象。     

The Effectiveness of Structural Monetary Policy and Macro-Prudential Policies — Based on the DSGE Model That Includes Bank Heterogeneous Credit

Journal of Systems Science and Complexity - Structural monetary policy and macro-prudential policies are important parts of the policy system of the People’s Bank of China. By constructing a...     

基于科技查新数据的用户拓展及其关联聚类

科技查新用户数据是查新业务积累的真实而宝贵的资源,其充分利用不仅是查新业务管理的需要,可提升查新服务水平,同时还能用于挖掘更多的情报价值.调研发现,当前对科技查新用户数据的挖掘利用仍以基础性的业务管理分析为主,而增值情报价值的挖掘较少.为拓展成都文献情报中心科技查新客户群体和增值服务范围,做大做强查新业务,立足成都中心查新服务实践,依托于积累查新用户及其服务记录,在挖掘发现潜在服务对象的基础上,通过多维用户关联聚类,探讨了后期信息情报服务的精准营销与推送,分析了基于用户聚类的服务推荐存在的不足,展望了挖掘更多需求信息的推荐路径.     

多维度深海鱼类识别算法

针对深海光线分布不均匀导致鱼类识别检测困难的问题,提出了符合视觉认知的多维度深海鱼类识别算法.该方法从时间维度优化传统的高斯混合模型(GMM)初步确定变化区域,从空间维度构建目标特征,完整提取运动目标,从时空关联维度建立深度学习的鱼类识别框架,试验结果表明:本算法可在多种复杂条件下准确提取运动目标,面积交迭度(AOM)达到80％以上,优于当前主流算法.     

飞秒光纤光源的波分解复用 复用实验

为研究飞秒光源作为链路初始光源的可行性，使用阵列波导光栅（AWG）对265 fs光纤光源进行波分解复用 复用实验。实验光源波长为1 52765～1 56505 nm，通道数为48，通道间隔为080 nm。研究发现:在分波实验中，飞秒光纤光源经过AWG后，扫描谱线3 dB带宽为064 nm ，相邻通道串扰为2726 dB，比连续光源增大502 dB；在合波实验中，观察到AWG多端口端对宽带光源的波长选择作用，验证了AWG的色散特性，发现多光束的输出谱线总体轮廓未出现太大变形；在波分解复用 复用实验中，谱线最大损耗为251 dB，输出谱线轮廓和初始光谱基本一致，信号在传输过程中未出现明显失真，证明飞秒光源作为链路初始光源的可行性。     

化学实验教学创新设计的调查报告

为了了解化学实验教学创新设计的现状,对常德市普通高中化学教师进行问卷调查,调查结果显示高中化学实验教学创新设计有弱化的趋势.在此基础上,从学科教学的角度对弱化的原因进行了分析,并提出化学实验教学创新设计的OCPP评价模式,尽可能建构、揭示或阐述实验教学创新设计的基本原理.OCPP模式包括目标评价(Objective Evaluation)、内容评价(Content Evaluation)、过程评价(Process Evaluation)和成果评价(Product Evaluation)4个部分.     

科技成果转化的探索与思考 ——基于湖南省科技成果转化的实证研究

科技成果转化是推动科技与经济结合的直接动力,是促进经济社会高质量发展的必要途径.文章通过对科技成果转化的重要性、现状和障碍的分析,提出科技行政管理部门应采取的管理举措.     

Evolutionary game analysis of low-carbon behavior credit supervision of logistics enterprises

Based on the low-carbon obligation fulfillment of Chinese logistics enterprises, this study constructs a tripartite evolutionary game model to analyze the evolutionary process of the interaction between the local government, logistics enterprises and the public in the process of low-carbon behavior credit supervision. Then using Netlogo software, a parameter simulation experiment is conducted to determine the optimal policy for improving the effect of supervision. The results are as follows: ① The combined influence of the local government and the public can effectively change the strategic choice of enterprises and promote the low-carbon behavior of enterprises. ② In terms of improving the effect of supervision, reducing the cost of government supervision would have a highly significant effect, and reducing the cost of the public would be counterproductive. ③ Increasing the government subsidies to enterprises and the government’s fines to enterprises both have a significant effect, and the effect of improving the former is better. However, increasing the severity of higher-level governments punishing local governments will reduce the stability of the system. ④ Supervision can be more effective by increasing the public’s impact on enterprises’ earnings rather than by increasing government subsidies to the public.