基于贝叶斯定理的垃圾邮件分类技术研究

邮件过滤技术是反垃圾邮件的重要手段,目前对垃圾邮件的过滤主要有基于内容、基于IP地址和基于信头、信封等方法,这些方法对垃圾邮件的过滤起到了一定作用.但是由于信体是垃圾邮件的最终载体,而仅依据IP地址、信头、信封中的特征容易造成错误判断.在贝叶斯分类器的工作原理的基础上,提出了基于贝叶斯分类器的反垃圾邮件模型的原理与实现方法,将反映垃圾邮件的特征综合在一起统称为"属性",避免了单纯基于IP、信头、信封过滤的规则性太强的缺点,降低将正常邮件判断为垃圾邮件的风险.     

基于朴素贝叶斯和层次聚类的两阶段垃圾邮件过滤方法

为降低对合法邮件的误判，提出一种基于朴素贝叶斯和层次聚类的两阶段垃圾邮件过滤方法。该方法将邮件划分为“合法邮件”、“可疑邮件”和“垃圾邮件”3类，在第一阶段，利用朴素贝叶斯算法速度快、分类性能好的优点，对邮件进行初步分类；在第二阶段，基于垃圾邮件的发送特征，利用层次聚类算法进行相似性比较。实验表明，该方法能够显著提高垃圾邮件的查准率，降低对合法邮件的误判，更加符合实际应用需求。     

Prioritized e‐mail servicing to reduce non‐spam delay and loss: A performance analysis

This paper proposes a prioritized e‐mail servicing on e‐mail servers to reduce the delay and loss of non‐spam e‐mails due to queuing. Using a prioritized two‐queue scheme, non‐spam e‐mails are queued in a fast queue and given higher service priority than spam e‐mails that are queued in a slow queue. Four prioritized e‐mail service strategies for the two‐queue scheme are proposed and analyzed. We modeled these four strategies using discrete‐time Markov chain analysis under different e‐mail traffic loads and service capacities. Non‐spam e‐mails can be delivered within a small delay, even under heavy e‐mail loadings and high spam‐to‐non‐spam a priori. Results from our analysis of the two‐queue scheme show that it gives non‐spam delay and loss probability two orders of magnitude smaller than the typical single‐queue approach during heavy spam traffic. Moreover, prioritized e‐mail servicing protects e‐mail servers from spam attacks. Copyright © 2007 John Wiley & Sons, Ltd.     

Packet level acknowledgement and Go‐Back‐N protocol performance in infrared wireless LANs

Infrared wireless LANs may employ repetition rate (RR) coding to increase the symbol capture probability at the receiver. This paper examines the effectiveness of RR coding to utilization for infrared LANs using the physical and link layer parameter values proposed in the Advanced Infrared (AIr) protocol standard, which is developed by the Infrared Data Association (IrDA). Infrared LANs employ a Go‐Back‐N (GBN) automatic repeat request (ARQ) retransmission scheme at the Link Control (LC) layer to ensure reliable information transfer. To efficiently implement RR coding, the receiver may return after every DATA packet a suggestion for the suitable RR value to be used by the transmitter and implement a Stop‐and‐Wait (SW) ARQ scheme at the medium access control (MAC) layer. The effectiveness of employing this optional SW ARQ scheme at the MAC layer is discussed. Analytical models for the ARQ retransmission schemes are developed and employed to compare protocol utilization for different link parameter values such as window size, packet length and LC time out periods. This analysis identifies the ARQ protocol that maximizes performance for the specific link quality and the implemented link layer parameters. The effectiveness of the proposed RR coding to LAN utilization for different ARQ scheme implementation is finally explored. This analysis identifies the link quality level at which RR should be adjusted for maximum performance. It is concluded that if the packet error rate is higher than 0.1–0.4 (depending on the implemented ARQ protocol), the receiver should advise the transmitter to double the implemented RR for maximum performance. These error rate values are high and can be effectively estimated by the transmitter based on packet retransmissions. Thus, the usefulness of the receiver indicating to the transmitter to adjust RR is questionable, as the transmitter can effectively implement the suitable RR value based on packet retransmissions. Copyright © 2003 John Wiley & Sons, Ltd.     

垃圾邮件及反垃圾邮件技术研究

当前,垃圾邮件日益泛滥,严重损害了互联网服务提供商、企业和消费者的利益。论文介绍了垃圾邮件的危害,并探讨了分别在邮件系统服务器端及客户端部署实施的各种反垃圾邮件技术和手段。     

Multi‐granular,multi‐purpose and multi‐Gb/s monitoring on off‐the‐shelf systems

As an attempt to make network managers’ life easier, we present M³Omon , a system architecture that helps to develop monitoring applications and perform network diagnosis. M³Omon behaves as an intermediate layer between the traffic and monitoring applications that provides advanced features, high performance and low cost. Such advanced features leverage a multi‐granular and multi‐purpose approach to the monitoring problem. Multi‐granular monitoring provides answers to tasks that use traffic aggregates to identify an event, and requires either flow records or packet data or even both to understand it and, eventually, take convenient countermeasures. M³Omon provides a simple API to access traffic simultaneously at several different granularities, i.e. packet‐level, flow‐level and aggregate statistics. The multi‐purposed design of M³Omon allows not only performing tasks in parallel that are specifically targeted to different traffic‐related purposes (e.g. traffic classification and intrusion detection) but also sharing granularities between applications, e.g. several concurrent applications fed from flow records that are provided by M³Omon . Finally, the low‐cost characteristic is brought by off‐the‐shelf systems (the combination of open‐source software and commodity hardware) and the high performance is achieved thanks to modifications in the standard NIC driver, low‐level hardware interaction, efficient memory management and programming optimization. Copyright © 2014 John Wiley & Sons, Ltd.     

图像垃圾邮件的发展和防范

垃圾邮件一直困扰着人们,图像垃圾邮件是垃圾邮件的变种,其危害远远大于传统的文本类垃圾邮件.本文首先给出了图像垃圾邮件的定义,介绍了图像垃圾邮件的类型和特点,然后分析了图像垃圾邮件的演进过程,总结了图像垃圾邮件的构造方法,最后探讨了图像垃圾邮件的防范措施.     

基于深度学习的垃圾邮件文本分类方法

文中旨在研究基于深度学习的垃圾邮件文本分类方法，该方法结合了卷积神经网络（CNN）和循环神经网络（RNN）的模型，通过对邮件文本进行特征提取和分类，能高效、准确地对垃圾邮件进行分类。文中以卷积神经网络和循环神经网络为实验对象，提出了一种垃圾邮件文本分类方法，并在公开数据集上进行了实验。实验结果表明，该方法在垃圾邮件文本分类任务上具有较高的准确率和召回率。     

基于兴趣相似度的协作式垃圾邮件过滤方案

由于电子邮件在互联网上的广泛应用,出现了垃圾邮件的泛滥和占用网络带宽,垃圾邮件给用户正常收发邮件造成了严重困扰。提出了一种新的基于兴趣相似度的协作式邮件过滤方案,使得兴趣相似的用户迅速聚合成信息共享的团体,每个用户在团体内选择出最合适的过滤规则集进行有效的共享。实验表明,该方案在保持一定的过滤精度的同时有效降低了误判率。     

Real‐time and fine‐grained network monitoring using in‐band network telemetry

In the modern era of software‐defined networking (SDN), network monitoring is becoming more important for providing information about a network and helping SDN controllers to make decisions about the network. In‐band Network Telemetry (INT) is a new network monitoring framework that collects packet‐level network information to provide real‐time and fine‐grained network monitoring. In this paper, we present the design of the overall INT management architecture and its two main components: the INT management system and INTCollector. The INT management system controls heterogeneous INT‐capable devices through a common interface. INTCollector is a high‐performance collector for INT data, which uses eXpress Data Path and an event detection mechanism. The evaluation result shows that INTCollector processes telemetry reports 27 times faster than other packet‐level telemetry collectors. We made the implementation as open source, to make researchers who are interested in INT implement their own ideas on top of our work.     

Dynamic Retry Adaptation Scheme to Improve Transmission of H.264 HD Video over 802.11 Peer‐to‐Peer Networks

This paper presents a dynamic retry adaptation scheme for H.264 HD video, called DRAS.264, which dynamically adjusts the retry limits of frames at the medium access control (MAC) layer according to the impact those frames have on the streamed H.264 HD video. DRAS.264 is further improved with a bandwidth estimation technique, better prediction of packet delays, and expanded results covering multi‐slice video. Our study is performed using the Open Evaluation Framework for Video Over Networks as a simulation environment for various congestion scenarios. Results show improvements in average peak signal‐to‐noise ratios of up to 4.45 dB for DRAS.264 in comparison to the default MAC layer operation. Furthermore, the ability of DRAS.264 to prioritize data of H.264 bitstreams reduces error propagation during video playback, leading to noticeable visual improvements.     

Deterministic process‐based generative models for characterizing packet‐level bursty error sequences

Errors encountered in digital wireless channels are not independent but rather form bursts or clusters. Error models aim to investigate the statistical properties of bursty error sequences at either packet level or bit level. Packet‐level error models are crucial to the design and performance evaluation of high‐layer wireless communication protocols. This paper proposes a general design procedure for a packet‐level generative model based on a sampled deterministic process with a threshold detector and two parallel mappers. In order to assess the proposed method, target packet error sequences are derived by computer simulations of a coded enhanced general packet radio service system. The target error sequences are compared with the generated error sequences from the deterministic process‐based generative model using some widely used burst error statistics, such as error‐free run distribution, error‐free burst distribution, error burst distribution, error cluster distribution, gap distribution, block error probability distribution, block burst probability distribution, packet error correlation function, normalized covariance function, gap correlation function, and multigap distribution. The deterministic process‐based generative model is observed to outperform the widely used Markov models. Copyright © 2013 John Wiley & Sons, Ltd.     

Cross‐layer optimization for CDMA/TDD wireless mesh networks

This paper proposes a new cross‐layer optimization algorithm for wireless mesh networks (WMNs). CDMA/TDD (code division multiple access/time division duplex) is utilized and a couple of TDD timeslot scheduling schemes are proposed for the mesh network backbone. Cross‐layer optimization involves simultaneous consideration of the signal to interference‐plus‐noise ratio (SINR) at the physical layer, traffic load estimation and allocation at medium access control (MAC) layer, and routing decision at the network layer. Adaptive antennas are utilized by the wireless mesh routers to take advantage of directional beamforming. The optimization formulation is subject to routing constraints and can be solved by general nonlinear optimization techniques. Comparisons are made with respect to the classic shortest‐path routing algorithm in the network layer. The results reveal that the average end‐to‐end successful packet rate (SPR) can be significantly improved by the cross‐layer approach. The corresponding optimized routing decisions are able to reduce the traffic congestion. Copyright © 2010 John Wiley & Sons, Ltd.     

基于文本分类技术的垃圾邮件识别系统

文章介绍了一个基于文本分类技术识别垃圾邮件系统的体系结构，并介绍了该系统涉及到的中文信息处理、文本特征选取、朴素贝叶斯分类器等关键技术。最后，文章给出了针对部分垃圾邮件的处理结果。结果表明，该方法对于垃圾邮件的识别，具有较好的效果。     

An Energy Saving Scheme for Multilane‐Based High‐Speed Ethernet

In this paper, we propose a scheme for partially dynamic lane control for energy saving in multilane‐based high‐speed Ethernet. In this scheme, among the given transmission lanes, at least one lane is always operating, and the remaining lanes are dynamically activated to alleviate the network performance in terms of queuing delay and packet loss in the range of acceptance. The number of active lanes is determined by the decision algorithm based on the information regarding traffic and queue status. The reconciliation sublayer adjusts the transmission lane with the updated number of lanes received from the algorithm, which guarantees no processing delay in the media access control layer, no overhead, and minimal delay of the exchanging control frames. The proposed scheme is simulated in terms of queuing delay, packet loss rate, lane changes, and energy saving using an OPNET simulator. Our results indicate that energy savings of around 55% (or, when the offered load is less than 0.25, a significant additional savings of up to 75%) can be obtained with a queuing delay of less than 1 ms, a packet loss of less than 10^?4, and a control packet exchange time of less than 0.5 μs in random traffic.     

Performance evaluation for single‐ and semi‐layer multistage interconnection networks servicing multicast traffic by full multicast operation

Here a novel analytical method for performance prediction estimation of single‐ and multi‐layer multistage interconnection networks (MINs) under multicast environments is presented. The ‘Cell Replication While Routing’ is used as a packet routing technique and the ‘full multicast’ mode as transmission policy is employed in all the MINs under study. This model is developed for single‐layer MINs and is extended and applied to special types of multi‐layer MINs, called semi‐layer MINs, which support mixed traffic patterns such as unicast and multicast in uniform traffic conditions. The results confirm the significant performance saturation of single‐layer MINs and the improvement of corresponding performance indicators by semi‐layer MINs. The proposed analytical framework is anticipated to be a very useful tool in evaluating the performance of semi‐layer MINs in designing more efficient QoS networks. Copyright © 2010 John Wiley & Sons, Ltd.     

Self‐adaptive clock frequency recovery based on buffer control

With high‐speed packet networks and packet transport networks (PTNs) beginning to support transmissions of high‐speed time division multiplex and multiplexer (TDM) and 3G services, the clock recovery in packet networks becomes significant and challengeable. This paper proposes a self‐adaptive clock frequency recovery based on buffer control to support the clock recovery in packet networks. It includes an estimation part and an amend part, estimates clock frequencies using the dynamic moving average algorithm, amends clock frequencies according to feedbacks, and balances stability, convergence, and system faults. Simulations show that the method has the ability of self‐adaption, avoids the buffer overflow and underflow, and achieves better accuracy, less packet delay, as well as better convergence ability. Copyright © 2011 John Wiley & Sons, Ltd.     

Terminal‐Assisted Hybrid MAC Protocol for Differentiated QoS Guarantee in TDMA‐Based Broadband Access Networks

This paper presents a terminal‐assisted frame‐based packet reservation multiple access (TAF‐PRMA) protocol, which optimizes random access control between heterogeneous traffic aiming at more efficient voice/data integrated services in dynamic reservation TDMA‐based broadband access networks. In order to achieve a differentiated quality‐of‐service (QoS) guarantee for individual service plus maximal system resource utilization, TAF‐PRMA independently controls the random access parameters such as the lengths of the access regions dedicated to respective service traffic and the corresponding permission probabilities, on a frame‐by‐frame basis. In addition, we have adopted a terminal‐assisted random access mechanism where the voice terminal readjusts a global permission probability from the central controller in order to handle the ‘fair access’ issue resulting from distributed queuing problems inherent in the access network. Our extensive simulation results indicate that TAF‐PRMA achieves significant improvements in terms of voice capacity, delay, and fairness over most of the existing medium access control (MAC) schemes for integrated services.     

Cross‐layer packet scheduler for QoS support over Digital Video Broadcasting‐Second Generation broadband satellite systems

This article presents a cross‐layer packet scheduler to provide end‐to‐end QoS guarantees for Digital Video Broadcasting‐Second Generation (DVB‐S2) broadband satellite systems. The scheduler design is based on a cross‐layer mechanism defined between the physical and the network layer. It includes an algorithm to guarantee the required QoS specifications established in the service level agreement. The algorithm calculation depends basically on two parameters: the available bandwidth present in a DVB‐S2 satellite link and the QoS requirements of each traffic class defined by the satellite operator. The cross‐layer scheduler's operation is demonstrated using the NS‐2 simulator tool. The results show that the proposed mechanism maximizes the bandwidth utilization while enforcing the priority level of each service class when an extreme reduction of bandwidth caused by rain events is experienced. Copyright © 2012 John Wiley & Sons, Ltd.     

Downlink media streaming with wireless fountain coding in wireline‐cum‐WiFi networks

This paper addresses the problem of streaming packetized media data in a combined wireline/802.11 network. Since the wireless channel is normally the bottleneck for media streaming in such a network, we propose that wireless fountain coding (WFC) be used over the wireless downlink in order to efficiently utilize the wireless bandwidth and exploit the broadcast nature of the channel. Forward error correction (FEC) is also used to combat errors at the application‐layer. We analytically obtain the moment generating function (MGF) for the wireless link‐layer delay incurred by WFC. With the MGF, the expected value of this wireless link‐layer delay is found and used by the access point (AP), who has no knowledge of the buffer contents of wireless receivers, to make a coding‐based decision. We then derive the end‐to‐end packet loss/late probability based on the MGF. We develop an integrated ns‐3/EvalVid simulator to evaluate our proposed system and compare it with the traditional 802.11e scheme which is without WFC capability but equipped with application‐ and link‐layer retransmission mechanisms. Through extensive simulations of video streaming, we show that streaming with WFC is able to support more concurrent video flows compared to the traditional scheme. When the deadlines imposed on video packets are relatively stringent, streaming with WFC also shows superior performance in terms of packet loss/late probability, video distortion, and video frame delay, over the traditional scheme. Copyright © 2011 John Wiley & Sons, Ltd.