从我做起，从我开始——《版权，我懂——中学生版权保护主题教育读本》读后

1994年我国引进第一部进口“大片”《亡命天涯》，是第一部与外商采用票房分账形式发行的，普通老百姓对“进口分账”的理解就是“大片”。这时人们似乎才知道在这之前，遍布街头巷尾的录像厅，通过关系才能看到的“内参片”放的就是这些“大片”。进口“大片”一定程度上刺激了国内电影市场的繁荣，在国内电影界和广大观众中产生了很大反响。其中一个现象就是：去电影院看，要花好几十块钱的票，而去街边小摊花几块钱买个碟就能先睹为快。

数字贸易视域下直播平台国际版权义务规则新趋势及我国因应

伴随全球数字贸易发展,网络直播突破地域性,直播中的版权侵权情况激增,问题也更为复杂,然而目前国际多边规则对直播平台版权治理缺乏明晰规制,但是区域性平台版权治理规则却不断繁荣,从ACTA到USMCA等区域性规则以及欧盟《版权指令》和美国DMCA现代化变革规则赋予了网络服务提供者高低不同的版权注意义务,义务规则呈现出“碎片化”及“强趋势”的特点,并且带来了一系列法益失衡的风险以及对新兴国家直播产业发展的抑制后果。当下我国正值直播产业“出海”占领国际市场的重要机遇,针对这种区域性规则的新趋势及其不利影响,我国首先应当破解“碎片化”,理顺当下国际规则中版权义务的核心规则,包括“红旗规则”下一般理性人应知的注意义务、“通知—屏蔽规则”下面向未来审查的注意义务、“一般过滤规则”下有义务预知的注意义务,且这三者呈现出梯度化的义务要求,义务程度逐级增加;其次应当对抗“强趋势”,划清注意义务的合理边界,减轻直播平台的不合理版权注意义务负担,结合直播平台特殊性构建直播平台版权注意义务规则;最后应当以国内法为基础,通过双边协定等带动区域逐步扩大我国规则的影响力,应对欧美区域性平台版权义务规则的扩散,主导具有中国特色的诸边体制,以对产业发展起到良性促进作用。

版权的教育豁免应明确立法

如何解决教育界面对的版权问题,是刚结束的检讨《版权条例》的若干条文咨询的一个重点。由于政府官员对教育豁免的认识及交代不足,使教育工作者对有关修订仍有很多不必要的忧虑,甚至大学方面,最近也赶紧在咨询完结前要求放宽《版权条例》。

WEIGHTED ALGORITHM FOR FINGERPRINTING RELATIONAL DATABASES

The necessity and the feasibility of introducing attribute weight into digital fingerprinting system are given. The weighted algorithm for fingerprinting relational databases of traitor tracing is proposed. Higher weights are assigned to more significant attributes, so important attributes are more frequently fingerprinted than other ones. Finally, the robustness of the proposed algorithm, such as performance against collusion attacks, is analyzed. Experimental results prove the superiority of the algorithm.

基于强化学习的水下高速航行体纵向运动控制研究

水下高速航行体由于空泡特性导致其数学模型存在强非线性和强不确定性,经典控制方法如线性二次型调节控制(linear quadratic regulator, LQR)、切换控制等很难实现有效控制。针对水下高速航行体模型难以准确解耦或线性化处理;经典控制方法难以充分考虑水下环境复杂多变性以及在应对扰动时控制器可能会出现过饱和现象的问题,采用智能控制中的强化学习算法,使用在不基于准确模型的条件下与环境不断探索与交互得到控制策略的策略,完成了深度确定性策略梯度(deep deterministic policy gradient,DDPG)智能体控制器的设计。实验结果证明,设计的控制器能够保证水下高速航行体纵向运动的稳定控制,在执行器不超过饱和范围内能够应对扰动并完成下潜控制任务,具有较强的鲁棒性和更好的适应性。

面向模糊C均值算法的MAME聚类有效性指标

聚类有效性指标可用来评估聚类结果的有效性,并且帮助判别聚类的类别数。现有的面向模糊C均值算法的聚类有效性指标存在对于类内紧致性的刻画不太到位、对于类间分离性的度量刻画不够准确的问题。为此,基于类内紧致性和类间分离性两个角度着手设计,提出了一种新的模糊聚类有效性指标——考虑最大值和均值的指标(maximum-mean,MAME)。首先,考虑了整个数据集的综合特征,计算分别分为K类和1类的情况的比值,提出了一种新的模糊紧致性度量表达式。其次,引入最大聚类中心距离和平均聚类中心距离,提出了一种新的分离性度量方法。最后,从模糊紧致性度量表达式、分离性度量方法出发,提出了MAME指标。面向5个UCI数据集和6个人工数据集,和9个聚类有效性指标(包括CH、DB、NPC、PE、FSI、XBI、NPE、WLI和I指标)一起进行了对比实验,验证了所提指标的准确性、稳定性,说明了MAME指标的鲁棒性较好。

融合聚类和小生境搜索的多模态多目标优化算法

针对多模态多目标优化中种群多样性难以维持和所得等价Pareto最优解数量不足问题,提出一种融合聚类和小生境搜索的多模态多目标优化算法(multimodal multi-objective optimization algorithm with clustering and niching searching,CSSMPIO)。首先利用基于聚类的特殊拥挤距离非支配排序方法(clustering-based special crowding distance,CSCD)初始化种群;引入自适应物种形成策略生成稳定的小生境,在不同的小生境子空间并行搜索和保持等价Pareto最优解;采用特殊拥挤距离非支配排序策略实现个体选优、精英学习策略避免过早收敛。通过在14个多模态多目标函数上进行测试,并与7种新提出的多模态多目标优化算法进行对比实验以及Wilcoxon秩和检验发现,CSSMPIO的总体性能优于对比算法。最后将算法用于基于地图的测试问题,进一步证明了算法的有效性。

深度残差收缩网络的多特征锅炉炉管声波信号故障识别

为了提高锅炉炉管声波信号故障识别的学习效果和识别精度,采用特征向量并行和拼接两种融合方式构成特征层,以及平均得分和最大值得分两种融合方式构建决策层等不同信息融合机制,提出基于深度残差收缩网络的多特征锅炉炉管声波信号故障识别方法。首先,考虑锅炉炉管上各声波传感器的差异性,分别计算声波信号谱特征一阶和二阶差分构建三通道特征集作为二维网络的输入特征向量;然后,在卷积神经网络和双向长短时记忆网络基础上引入注意力机制构建基线模型,并采用深度残差收缩网络对二维网络的通道权重进行优化分配,提高模型的故障识别精度。大量实验结果表明:采用特征向量并行融合方式构成特征层的信息融合机制是一种更有效的策略;本文模型的识别精度得到较大程度提高,与基线模型相比较,未加权平均召回率提高了4.32%。

优化分类的弱目标孪生网络跟踪研究

针对传统孪生网络算法对模糊、低分辨率等弱目标跟踪效果不佳的问题,提出了优化分类预测的孪生网络算法。首先通过引入可变形卷积模块,提高骨干网络特征提取能力,其次在分类分支中引入位置信息,提升算法对于目标的识别能力,最后使用轻量级的卷积神经网络进行分类预测和边界预测任务,在规避多尺度测试的同时,进一步利用了图像的语义信息,使跟踪结果具有较高的可信度。在OTB2015、VOT2018公共数据集上进行的大量实验表明,本文算法综合表现优于主流同类算法,对模糊、形变、快速运动等多种复杂场景具有较好的适应性。

数据流形边界及其分布条件的增量式降维算法

为了解决增量流形学习中的噪声干扰,以及对不同分布状态下的新数据进行流形降维问题,本文提出一种数据流形边界及其分布条件的增量式降维算法(incremental dimensionality reduction algorithm based on data manifold boundaries and distribution state,IDR-DMBDS)。该算法首先分析噪声概率分布同时对数据降噪,确定降噪数据的流形形态为主流形,并在主流形上表征出噪声的分布形式,以此获得近似的原数据流形边界,然后基于流形边界判别新数据的分布状态,最后将分布于原流形形态之上以及之外的新数据分别映射至低维空间。实验表明,该算法能够有效实现基于流形的增量式高维含噪数据的低维特征挖掘。

基于点集匹配的缺陷样本图像生成方法

针对工业缺陷检测中存在的由产品次品率过低、产品迭代更新过快、缺陷种类难以覆盖全部以及缺陷样本高质量标注难度较高导致的小样本问题,使用基于点集匹配的缺陷样本图像生成方法来对缺陷数据进行扩充。将缺陷部位从多特征角度进行变换,使用单张样本进行扩充得到不同特征的缺陷图像,解决小样本条件下深度学习方法难以生成高质量缺陷图像的问题。通过图像评估与实验验证,该方法生成的图像具有更好的视觉效果,并且对缺陷与分割模型有着高效的提升。该方法可应用于样本较少的深度学习模型训练过程中,达到扩充样本提高训练效果的目的。

基于空时对抗变分自编码器的人群异常行为检测

基于视频的人群异常行为检测对提前发现安全风险、预防群体安全事故发生具有重要价值。针对人群异常行为事件的稀少性导致的无法直接充分学习异常样本的表示、异常事件检测精度低的问题,在变分自编码器基础上,提出一种基于预测的空时对抗变分自编码器(spatial-temporal adversarial variational autoencoder,STAVAE)视频异常检测模型,通过引入长短期记忆网络(long short-term memory,LSTM)和对抗网络模块,对正常样本视频序列的时间维度与空间维度进行联合特征表示与重构,减少了正常样本重建过程中的特征损失进而扩大了异常样本的预测损失,避免了对异常样本的依赖,实现了基于模型重构误差的人群逃散异常行为检测。在公开数据集UMN及采集视频数据集上进行对比实验,证明ST-AVAE模型在基于监控视频的人群异常逃散行为检测中均具有最优的检测精度和召回率,对抗网络模块显著提升了异常检测的性能。

GEOMETRICALLY INVARIANT WATERMARKING BASED ON RADON TRANSFORMATION

The weakness of classical watermarking methods is the vulnerability to geometrical distortions that widely occur during normal use of the media. In this letter, a new image- watermarking method is presented to resist Rotation, Scale and Translation (RST) attacks. The watermark is embedded into a domain obtained by taking Radon transform of a circular area selected from the original image, and then extracting Two-Dimensional (2-D) Fourier magnitude of the Radon transformed image. Furthermore, to prevent the watermarked image from degrading due to inverse Radon transform, watermark signal is inversely Radon transformed individually. Experimental results demonstrate that the proposed scheme is able to withstand a variety of attacks including common geometric attacks.

NEW APPROACH TO COPYRIGHT PROTECTION OF IMAGE AGAINST GEOMETRICAL ATTACKS

To protect the copyright of the image as well as the image quality, a kind of image zero-watermark method based on the Krawtchouk moment invariants and timestamp is proposed. A method is used to protect the image, in which features are drawn out from the image as the watermarking. The main steps of the method are presented. Firstly, some low-order moment invariants of the image are extracted. Secondly, the moment invariants and the key are registered to a fair third party to gain the timestamp. Finally, the timestamp can be used to prove who the real owner is. The processing method is simple, only with a few low-order moment invariants to be computed. Experimental results are obtained and compared with those of the method based on geometric moment invariants. Results show that the scheme can well withstand such geometrical attacks as rotating, scaling, cropping, combined attack, translating, removing lines, filtering, and JPEG lossy compression.

Digital Rights Management：Model,Technology and Application

with rapid achievement of current information technology and computing ability and applications,much more digital content such as films,cartoons,design drawings,office documents and software source codes are produced in daily work,however to protect the content being copying,shared or deliberately stolen by inside or outside,digital rights management(DRM) became more and more important for digital content protection.In this paper,we studied various DRM model,technology and application,and first proposed DRM Security Infrastructure(DSI),in which we defined encryption,hash,signature algorithm,watermarking algorithms,authentication,usage control,trusted counter,conditional trace,secure payment,and based on the DSI we then proposed a whole classification approach and architecture of all kinds of DRMs,in which we proposed 6 typical classes of copyrights and content protection DRMs architecture:(1) Software-oriented DRM,(2) e Book-oriented DRM,(3) Video-oriented DRM,(4) Image-Oriented DRM(5) Unstructured data oriented DRM,(6) Text-oriented DRM.Based on the above DSI,we then proposed a dynamic DRM model selection method for various DRM application,which can be adapted dynamically for different technology of different applications,which can provide awhole solution for variant DRM development in a rapid and customized mode.The proposed DRM method,technology and application in this paper provided a common,flexible and extendable solution for variant DRM scenes,and can support rapid and customized development.Moreover,we proposed an opinion that the future life will enter into a new era that the content usage and consumption will not again adopt DRM technology rather than with law,liberty and morality.

An Audio Zero-Watermark Scheme Based on Energy Comparing

Zero-watermark technique,embedding watermark without modifying carriers,has been broadly applied for copyright protection of images.However,there is little research on audio zerowatermark.This paper proposes an audio zero-watermark scheme based on energy relationship between adjacent audio sections.Taking use of discrete wavelet transformation(DWT),it gets power approximations,or energies,of audio segments.Then,it extracts the audio profile,i.e.the zero-watermark,according to the relative size of energies of consecutive fragments.The experimental results demonstrate that the proposed scheme is robust against general malicious attacks including noise addition,resampling,low-pass filtering,etc.,and this approach effectively solves the contradiction between inaudibility and robustness.

Multipurpose audio watermarking algorithm

To make audio watermarking accomplish both copyright protection and content authentication with localization, a novel multipurpose audio watermarking scheme is proposed in this paper. The zero-watermarking idea is introduced into the design of robust watermarking algorithm to ensure the transparency and to avoid the interference between the robust watermark and the semi-fragile watermark. The property of natural audio that the VQ indices of DWT-DCT coefficients among neighboring frames tend to be very similar is utilized to extract essential feature from the host audio, which is then used for watermark ex-traction. And, the chaotic mapping based semi-fragile watermark is embedded in the detail wavelet coefficients based on the instantaneous mixing model of the independent component analysis (ICA) system. Both the robust and semi-fragile watermarks can be extracted blindly and the semi-fragile watermarking algorithm can localize the tampering accurately. Simulation results demonstrate the effectiveness of our algorithm in terms of transparency, security, robustness and tampering localization ability.

Attribute Based DRM Scheme with Dynamic Usage Control in Cloud Computing

In order to achieve fine-grained access control in cloud computing,existing digital rights management(DRM) schemes adopt attribute-based encryption as the main encryption primitive.However,these schemes suffer from inefficiency and cannot support dynamic updating of usage rights stored in the cloud.In this paper,we propose a novel DRM scheme with secure key management and dynamic usage control in cloud computing.We present a secure key management mechanism based on attribute-based encryption and proxy re-encryption.Only the users whose attributes satisfy the access policy of the encrypted content and who have effective usage rights can be able to recover the content encryption key and further decrypt the content.The attribute based mechanism allows the content provider to selectively provide fine-grained access control of contents among a set of users,and also enables the license server to implement immediate attribute and user revocation.Moreover,our scheme supports privacy-preserving dynamic usage control based on additive homomorphic encryption,which allows the license server in the cloud to update the users' usage rights dynamically without disclosing the plaintext.Extensive analytical results indicate that our proposed scheme is secure and efficient.