A privacy data leakage prevention method in P2P networks

Registered peers to a P2P service can share and exchange information with other peers without servers using P2P network. In such P2P networks, there are frequent leaks of the internal privacy data of an organization through P2P file sharing. Today, DLP which is a privacy data leakage prevention technology is applied P2P network blocking and file encryption methods. However restricting all the services and normal users is difficult due to the number of ports used by P2P including the port 80. Thus, we propose a privacy data leakage prevention method by releasing a P2P sharing file that does not include privacy data using a privacy data removing technology with a privacy data leaking risk factor. The proposed method provides higher security and performance compared with a DLP method as privacy data is removed from a P2P sharing file.     

Intelligent-memory architecture for artificial neural networks

Execution of artificial neural networks, especially for online pattern recognition, mainly depends on time-efficient execution of weighted sums. A new architecture achieves this goal, with a computation time superior to the time complexity of sequential von Neumann machines. This architecture uses additional logic to extend the functionality of conventional RAM. The authors discuss an implementation of this architecture that uses reconfigurable logic     

卷积神经网络的人脸隐私保护识别

目的 近年来，随着人脸识别认证技术的发展及逐渐普及，大量人脸照片存放在第三方服务器上的现象十分普遍，如何对人脸进行隐私保护这个问题变得十分突出。方法 首先对人脸图像进行预处理，然后采用Arnold变换对人脸关键部位进行分块随机置乱，并将置乱结果图输入到深度卷积神经网络中。为了解决人脸照片在分块置乱时由于本身拍照角度的原因导致的分块不均等因素，在预处理时根据人眼进行特性点定位，再据此进行对齐处理，使得预处理后的照片人眼处于同一水平线。针对人脸隐私保护及加扰置乱后图像的识别，本文提出了基于分块随机加扰的深度卷积神经网络模型。不包含附加层，该模型网络结构由4个卷积层、3个池化层、1个全连接层和1个softmax回归层组成。服务器端通过深度神经网络模型直接对置乱后人脸图像进行验证识别。结果 该算法使服务器端全程不存储原始人脸模板，实现了对原始人脸图像的有效加扰保护。实验采用该T深度卷积神经网络对处理过后的ORL人脸库进行识别，最终识别准确率达到97.62%。同时通过多组对比实验，验证了本文方法的有效性。结论 与其他文献中手工提取特征并利用决策树和随机森林进行训练识别的方法相比，本文方法减少了人工提取特征的工作量，且具有高识别率。     

A new winners-take-all architecture in artificial neural networks

MAXNET is a common competitive architecture to select the maximum or minimum from a set of data. However, there are two major problems with the MAXNET. The first problem is its slow convergence rate if all the data have nearly the same value. The second one is that it fails when either nonunique extreme values exist or each initial value is smaller than or equal to the sum of initial inhibitions from other nodes. In this paper, a novel neural network model called SELECTRON is proposed to select the maxima or minima from a set of data. This model is able to select all the maxima or minima via competition among the processing units even when MAXNET fails. We then prove that SELECTRON converges to the correct state in every situation. In addition, the convergence rates of SELECTRON for three special data distributions are derived. Finally, simulation results indicate that SELECTRON converges much faster than MAXNET.     

Scalable network-on-chip architecture for configurable neural networks

Providing highly flexible connectivity is a major architectural challenge for hardware implementation of reconfigurable neural networks. We perform an analytical evaluation and comparison of different configurable interconnect architectures (mesh NoC, tree, shared bus and point-to-point) emulating variants of two neural network topologies (having full and random configurable connectivity). We derive analytical expressions and asymptotic limits for performance (in terms of bandwidth) and cost (in terms of area and power) of the interconnect architectures considering three communication methods (unicast, multicast and broadcast). It is shown that multicast mesh NoC provides the highest performance/cost ratio and consequently it is the most suitable interconnect architecture for configurable neural network implementation. Routing table size requirements and their impact on scalability were analyzed. Modular hierarchical architecture based on multicast mesh NoC is proposed to allow large scale neural networks emulation. Simulation results successfully validate the analytical models and the asymptotic behavior of the network as a function of its size.     

Deep neural networks architecture driven by problem-specific information

Neural Computing and Applications - Deep learning provides a variety of neural network-based models, known as deep neural networks (DNNs), which are being successfully used in several domains to...     

Polynomial neural networks architecture: analysis and design

In this study, we introduce and investigate a class of neural architectures of Polynomial Neural Networks (PNNs), discuss a comprehensive design methodology and carry out a series of numeric experiments. Two kinds of PNN architectures, namely a basic PNN and a modified PNN architecture are discussed. Each of them comes with two types such as the generic and the advanced type. The essence of the design procedure dwells on the Group Method of Data Handling. PNN is a flexible neural architecture whose structure is developed through learning. In particular, the number of layers of the PNN is not fixed in advance but becomes dynamically meaning that the network grows over the training period. In this sense, PNN is a self-organizing network. A comparative analysis shows that the proposed PNN are models with higher accuracy than other fuzzy models.     

具有漏泄时滞的高阶细胞神经网络的多周期性分析

分析了一类具有漏泄时滞的高阶细胞神经网络的多周期性和指数收敛性. 给出了保证此类网络的周期环在饱和区内局部指数收敛的充分条件. 所得结果表明, 一个n维网络可以有2^n个周期环存在于饱和区, 而且这些周期环是局部指数收敛的. 仿真实例进一步证明了结论的有效性.     

具漏泄时滞的细胞神经网络多周期性分析

针对一类具有漏泄时滞细胞神经网络模型,首先给出该类网络的周期环在饱和区局部指数收敛的充分条件.研究表明,一个n维网络可以有2n个周期环存在于饱和区,并且这些周期环是局部指数收敛的.然后,研究了该时滞细胞神经网络指数周期的一个特殊情形--指数稳定.数值例子和仿真结果验证了所得结果的有效性.     

Layered tile architecture for efficient hardware spiking neural networks

Spiking Neural Network (SNN) is the most recent computational model that can emulate the behaviour of biological neuron system. However, its main drawback is that it is computationally intensive, which limits the system scalability. This paper highlights and discusses the importance and significance of emulating SNNs in hardware devices. A layer-level tile architecture (LTA) is proposed for hardware-based SNNs. The LTA employs a two-level sharing mechanism of computing components at the synapse and neuron levels, and achieves a trade-off between computational complexity and hardware resource costs. The LTA is implemented on a Xilinx FPGA device. Experimental results demonstrate that this approach is capable of scaling to large hardware-based SNNs.     

Sociotechnical architecture for online privacy

Users' concerns regarding privacy issues are lowering their trust in e-services and, thus, affecting the widespread adoption of online services. To increase users' perceived control over their privacy, the authors propose a novel e-privacy architecture.     

Internal leakage detection for feedwater heaters in power plants using neural networks

As interest in safety and performance of power plants becomes more serious and wide-ranging, the significance of research on turbine cycles has attracted more attention. This paper particularly focuses on thermal performance analysis under the conditions of internal leakages inside closed-type feedwater heaters (FWHs) and their diagnosis to identify the locations and to quantify leak rates. Internal leakage is regarded as flow movement through the isolated path but remaining inside the system boundary of a turbine cycle. For instance, leakages through the cracked tubes, tube-sheets, or pass partition plates in a FWH are internal leakages. Internal leakages impact not only plant efficiency, but also direct costs and/or even plant safety associated with the appropriate repairs. Some types of internal leakages are usually critical to get the parts fixed and back in a timely manner. The FWHs installed in a Korean standard nuclear power plant were investigated in this study. Three technical steps have been, then, conducted: (1) the detailed modeling of FWHs covering the leakage from tubes, tube-sheets, or pass partition plates using the simulation model, (2) thermal performance analysis under various leakage conditions, and (3) the development of a diagnosis model using a feed-forward neural network, which is the correlation between thermal performance indices and leakage conditions. Since the operational characteristics of FWHs are coupled with one another and/or with other neighbor components such as turbines or condensers, recognizing internal leakages is difficult with only an analytical model and instrumentation at the inlet and outlet of tube- and shell-sides. The proposed neural network-based correlation was successfully validated for test cases.     

A hybrid precision low power computing-in-memory architecture for neural networks

Recently, non-volatile memory-based computing-in-memory has been regarded as a promising competitor to ultra-low-power AI chips. Implementations based on both binarized (BIN) and multi-bit (MB) schemes are proposed for DNNs/CNNs. However, there are challenges in accuracy and power efficiency in the practical use of both schemes. This paper proposes a hybrid precision architecture and circuit-level techniques to overcome these challenges. According to measured experimental results, a test chip based on the proposed architecture achieves (1) from binarized weights and inputs up to 8-bit input, 5-bit weight, and 7-bit output, (2) an accuracy loss reduction of from 86% to 96% for multiple complex CNNs, and (3) a power efficiency of 2.15TOPS/W based on a 0.22μm CMOS process which greatly reduces costs compared to digital designs with similar power efficiency. With a more advanced process, the architecture can achieve a higher power efficiency. According to our estimation, a power efficiency of over 20TOPS/W can be achieved with a 55nm CMOS process.     

Application of convolutional neural networks for preventing information leakage in open internet resources

The architecture of convolutional neural networks has been considered, including the types of layers used and the principles of their operation, settings, and training features. The possibilities of applying this type of network to solve the problems of information leakage prevention in natural language have been described. The possibility of applying them to solve the problem of classifying Internet pages that serve as web resources to identify pages of interest has been studied.     

Predicting piezometric water level in dams via artificial neural networks

The safety control of dams is based on measurements of parameters of interest such as seepage flows, seepage water clarity, piezometric levels, water levels, pressures, deformations or movements, temperature variations, loading conditions, etc. Interpretation of these large sets of available data is very important for dam health monitoring and it is based on mathematical models. Modelling seepage through geological formations located near the dam site or dam bodies is a challenging task in dam engineering. The objective of this study is to develop a feedforward neural network (FNN) model to predict the piezometric water level in dams. An improved resilient propagation algorithm has been used to train the FNN. The measured data have been compared with the results of FNN models and multiple linear regression (MLR) models that have been widely used in analysis of the structural dam behaviour. The FNN and MLR models have been developed and tested using experimental data collected during 9 years. The results of this study show that FNN models can be a powerful and important tool which can be used to assess dams.     

A middleware architecture for privacy protection

The issue of user privacy is constantly in spotlight since an ever increasing number of online services collects and processes personal information from users, in the context of personalized service provision. In fact, recent advances in mobile communications, location and sensing technologies and data processing are boosting the deployment of context-aware personalized services and the creation of smart environments; but at the same time, they pose a serious risk on individuals’ privacy rights. Although technology makes the collection of data easy, its protection against abuse is left to data protection legislation. However, the privacy requirements, other than being general and abstract terms to be regarded as legislature issues, should be brought down in technological reality and carefully accounted for in devising technical solutions. In order to limit the disclosure and avoid the misuse of personal data, this paper discusses an architectural proposal for a middleware system that will enforce protection of user privacy through technical means. The proposed architecture mediates between the users, the service providers and the law, constituting a middleware shield for individuals’ personal data.     

Security and privacy in sensor networks

Sensor networks offer economically viable solutions for a variety of applications. For example, current implementations monitor factory instrumentation, pollution levels, freeway traffic, and the structural integrity of buildings. Other applications include climate sensing and control in office buildings and home environmental sensing systems for temperature, light, moisture, and motion. Sensor networks are key to the creation of smart spaces, which embed information technology in everyday home and work environments. The miniature wireless sensor nodes, or motes, developed from low-cost off-the-shelf components at the University of California, Berkeley, as part of its smart dust projects, establish a self-organizing sensor network when dispersed into an environment. The privacy and security issues posed by sensor networks represent a rich field of research problems. Improving network hardware and software may address many of the issues, but others will require new supporting technologies.     

传感器网络中的安全性和保密性

为了缓解传感器网络在安全性和保密性方面的困境,对传感器节点、窃听、监测数据、拒绝服务攻击等易被攻击环节的安全性问题进行了探讨,并从硬件和软件两个方面,提出了改进协议、数据加密、访问控制、多路发送、分布式询问等多种抵御攻击的办法。     

Computational data privacy in wireless networks

Wireless networks appeal to privacy eavesdroppers that secretly capture packets from the open medium and read the metadata and data content in search of any type of information through data mining and statistical analysis. Addressing the problems of personal privacy for wireless users will require flexible privacy protection mechanisms adaptively to the frequently varying context. Privacy quantification is prerequisite for enabling a context-aware privacy protection in wireless networks. This paper proposes to quantify the data privacy during wireless communication processes. We give a computational quantification method of data privacy through introducing the concepts of privacy entropy and privacy joint entropy, which permits users and applications to on-demand customize their preferential sensitivity extent of privacy leakage and protection strength. Accordingly, we put forward a data-privacy protection scheme in order to explain how to utilize the proposed computational quantification method of data privacy against the excessive disclosure of data privacy. The results show that the computational quantitation method can effectively characterize the real-time fluctuation of data privacy during communication processes and provide the reliable judgment to context-aware privacy protection, which enables to real-timely control the present data privacy to an anticipated target and to balance a tradeoff between the data privacy and communication efficiency.     

Pre-synaptic lateral inhibition provides a better architecture for self-organizing neural networks

Unsupervised learning is an important ability of the brain and of many artificial neural networks. A large variety of unsupervised learning algorithms have been proposed. This paper takes a different approach in considering the architecture of the neural network rather than the learning algorithm. It is shown that a self-organizing neural network architecture using pre-synaptic lateral inhibition enables a single learning algorithm to find distributed, local, and topological representations as appropriate to the structure of the input data received. It is argued that such an architecture not only has computational advantages but is a better model of cortical self-organization.