Peer-to-Peer+Traffic+Identification+Using+Bayesian+Networks

合集下载

Peer-To-Peer介绍

Peer-To-Peer介绍

Peer-To-Peer介绍最近几年,Peer-to-Peer (对等计算,简称P2P) 迅速成为计算机界关注的热门话题之一,财富杂志更将P2P列为影响Internet未来的四项科技之一。

“Peer”在英语里有“对等者”和“伙伴”的意义。

因此,从字面上,P2P可以理解为对等互联网。

国内的媒体一般将P2P翻译成“点对点”或者“端对端”,学术界则统一称为对等计算。

P2P可以定义为:网络的参与者共享他们所拥有的一部分硬件资源(处理能力、存储能力、网络连接能力、打印机等),这些共享资源通过网络提供服务和内容,能被其它对等节点(Peer)直接访问而无需经过中间实体。

在此网络中的参与者既是资源(服务和内容)提供者(Server),又是资源获取者(Client)。

客观地说,这种计算模式并不是什么新技术,自从上个世纪70年代网络产生以来就存在了,只不过当时的网络带宽和传播速度限制了这种计算模式的发展。

90年代末,随着高速互联网的普及、个人计算机计算和存储能力的提升,P2P技术重新登上历史舞台并且带来了一场技术上的革命。

许多基于P2P技术的杀手级应用应运而生,给人们的生活带来了极大的便利。

从计算模式上来说,P2P打破了传统的Client/Server (C/S)模式,在网络中的每个结点的地位都是对等的。

每个结点既充当服务器,为其他结点提供服务,同时也享用其他结点提供的服务。

P2P与C/S模式的对比如下图所示:图1 Client/Server模式图2 Peer to Peer 模式P2P技术的特点体现在以下几个方面[1]:•非中心化:网络中的资源和服务分散在所有结点上,信息的传输和服务的实现都直接在结点之间进行,可以无需中间环节和服务器的介入,避免了可能的瓶颈。

P2P的非中心化基本特点,带来了其在可扩展性、健壮性等方面的优势。

•可扩展性:在P2P网络中,随着用户的加入,不仅服务的需求增加了,系统整体的资源和服务能力也在同步地扩充,始终能比较容易地满足用户的需要。

p2p的原理

p2p的原理

p2p的原理P2P(Peer-to-Peer)技术是一种去中心化的网络通信模式,其原理是将网络节点平等对待,所有的节点不再依赖一个中央服务器来进行通信和数据传输。

在P2P网络中,每个节点既是客户端又是服务器,它们可以直接与其他节点进行通信,同时也可以充当中转站来传输数据。

节点之间通过建立连接,相互交换和共享资源,使得每个节点都可以充当网络的接入点和提供者。

P2P网络的通信过程主要分为三个步骤:1. 节点发现:每个节点加入网络后,需要通过某种方式发现其他的节点,并与之建立连接。

常见的方法包括使用中央服务器进行节点列表的维护,或是使用分布式哈希表(DHT)等算法来实现节点的自动发现。

2. 连接建立:节点之间通过协议来建立连接,常用的协议包括TCP(Transmission Control Protocol)和UDP(User Datagram Protocol)。

连接建立后,节点可以直接进行数据传输和通信,不需要经过中央服务器进行转发。

3. 数据传输:节点之间可以通过直接连接来传输数据,也可以通过中继节点进行转发。

数据在P2P网络中会被分割成多个小块,然后分布在不同的节点上。

当节点需要获取某个数据块时,它可以从其他节点请求该块并进行下载,直到将完整的数据下载完成。

由于P2P网络的去中心化特性,它具有较高的可靠性和弹性。

即使网络中的某些节点无法正常工作或离线,其他节点仍然可以继续运行和提供服务。

同时,P2P网络也具有较好的扩展性,可以支持大规模的节点加入并提供更强大的计算和存储能力。

总的来说,P2P网络通过节点之间的直接连接和资源共享,实现了无需中央服务器的通信和数据传输,从而提供了一种高效、可靠且灵活的网络通信模式。

上海市第三女子中学2021-2022学年高三下学期期中线上检测英语试卷

上海市第三女子中学2021-2022学年高三下学期期中线上检测英语试卷

2021学年第二学期高三年级英语学科线上教学评估120 minutes | 满分140分I. Listening Comprehension (25%)Section A (10分)Directions: In Section A, you will hear ten short conversations between two speakers. At the end ofeach conversation, a question will be asked about what was said. The conversations and the questionswill be spoken only once. After you hear a conversation and the question about it, read the fourpossible answers on your paper, and decide which one is the best answer to the question you haveheard.1. A. Manager and secretary. B. Customer and salesgirl. C. Man and wife. D. Salesman and manufacturer.2. A. $240. B. $290. C. $250. D. $200.3. A. At a museum. B. At a library. C. At a bookstore. D. At a coffee shop.4. A. Tonight. B. On Friday night. C. On Saturday. D. Next week.5. A. Take care of the baby. B. Fix the air-conditioner. C. Stay in the room alone. D. Get a repairman.6. A. There is no convenient store nearby. B. Convenient stores may not sell phone chargers.C. The phone charger sold online is of poor quality.D. Nobody will be available at the convenient store.7. A. His membership is no longer valid. B. He is tired of that gym.C. He never went to that gym.D. He went to that gym only once.8. A. He might be a perfectionist. B. He is arguing with his colleagues.C. He is annoyed by the constant changes.D. He is not bright enough to understand the plan.9. A. More courses benefit the woman. B. The woman will be worn out if she takes 6 courses.C. The woman will have to book a room in the library.D. The woman should live in the library for the exams.10. A. She can only cook Chinese-style dishes. B. She learned her cooking from her mother.C. It always takes her a long time to prepare a meal.D. She will order some food for her house-warming party. Section BDirections:In Section B, you will hear two short passages and one longer conversation, and you will be asked several questions on each of the passages and the longer conversation. The passages and the longer conversation will be read twice, but the questions will be spoken only once. When you hear a question, read the four possible answers on your paper and decide which one would be the best answer to the question you have heard.Questions 11 through 13 are based on the following passage.11. A. Bad weather. B. Mistakes made by the air traffic controllers.C. The loss of control of the airplane.D. The duty of the pilot.12. A. The pilot of the second plane made a big mistake.B. They avoided each other by turning in different directions.C. They narrowly escaped crashing into each other.D. One plane was suddenly out of control.13. A. Air controllers are often careless. B. The importance of the pilots.C. Air travel is much safer than other means of travelling.D. The potential danger of air travel.Questions 14 through 16 are based on the following passage.14. A. Parents have little control over what children are watching or playing. B. The media are full of violence.C. Children think that violence and crime are normal.D. Young people are harder to discipline.15. A. They should keep their kids from watching movies or playing games.B. They should ask for help from the schools.C. They should speak out against the entertainment industry.D. They should take their children out for walks.16. A. The differences between the past and the present. B. People’s attitudes toward violence in the media.C. How violence in the media affects society.D. The necessity of keeping children from playing video games. Questions 17 through 20 are based on the following conversation.17. A. Something is wrong with the engine. B. The car never went wrong before.C. The car is too old to drive.D. She is not sure about the problem.18. A. In Japan. B. In Germany. C. In America. D. In Italy.19. A. The mechanic charged her a reasonable price. B. The mechanic gave her an unreasonable price.C. The mechanic overcharged her.D. The mechanic undercharged her.20. A. The mechanic is undependable. B. The mechanic is tricky.C. The mechanic is trustworthy.D. The mechanic is helpful.Ⅱ. Grammar and Vocabulary (20%)Section ADirections:After reading the passage below, fill in the blanks to make the passage coherent and grammatically correct. For the blanks with a given word, fill in each blank with the proper form of the given word; for the other blanks, use one word that best fits each blank.Tea in BritainTea may be thought as the most typical of English drinks, but it is actually a relative latecomer to British shores. Although the custom of drinking tea (21)______ (date) back to the 3rd millennium (一千年) BC in China, the drink didn’t appear in England until the mid-17th century.Curiously, it is said that it was the London coffee houses (22)______ were responsible for introducing tea to England. One of the first coffee house merchants (23)______ (offer) tea was Thomas Garway, who sold both liquid and dry tea to the public as early as 1657. Three years later, he issued a big brochure advertising the virtues of tea, (24)______ its positive effect on health.Tea gained popularity quickly in the coffee houses, and this distressed the pub owners (25)______ tea cut their sales of beer. At the same time, it was also bad news for the government officials, (26)______ ______ many depended on a steady stream of revenue from taxes on liquor sales. By 1750, tea (27)______ (become) the favored drink of Britain’s lower classes.A 1676 act taxed tea and required coffee house operators to apply for a license. The heavy taxation had the effect of creating a whole new industry — tea smuggling (走私). Even smuggled tea was expensive, however, and therefore extremely profitable, so many smugglers began to adulterate (掺假) the tea with other substances, such as willow andsloe leaves. (28)______ (use) tea leaves were also redried and added to fresh leaves.Finally, 1784 William Pitt the Younger determined to take (29)______ measures that could reverse the situation. He introduced the Commutation Act, which dropped the tax on tea from 119% to 12.5%, effectively ending smuggling. He also took many other measures, hoping that such attempts (30)______ stop adulteration. But adulteration remained a problem until the Food and Drug Act of 1875 brought in severe punishment for the practice.Section BDirections:Complete the following passage by using the words in the box. Each word can only be used once. Note that there is one word more than you need.A. modestB. creativeC. confirmD. paceE. increasinglyF. emergingG. scoring H. caps I. evenly J. access K. collapseBathroom Reveals Your Repayment AbilityBanks and other lenders typically look at borrowers’ credit histories, tax forms and other financial information to determine whether they will get paid back. In China, (31)______ lenders also look at borrowers’ bathrooms.As the economy slows, the government wants to develop a credit culture to get Chinese families spending instead of saving. While the country is getting wealthier, the financial system has not kept (32)______. Some people have had no (33)______ to credit card or mortgage, so lenders often have little reliable information about potential borrowers.To fill the gaps, one lender, China Rapid Finance, assigns its investigators to check for the number of toothbrushes or towels to determine how many people are living in a house. They look for dirty dishes in the kitchen. They take photos of a potential borrower at work to (34)______ employment status.A growing number of companies are trying to crack the credit code in China. The Internet giants Alibaba, Tencent and Baidu are developing credit (35)______ systems based on users’ online transactions (交易) and search histories.The World Bank estimated that 79 percent of China’s population above the age of 15 have bank accounts, while only 10 percent have ever borrowed from the formal financial system. China’s banks have a(n) (36)______ consumers lending business. Contrarily, online lenders are (37)______ as the pioneers, satisfying China’s rising con sumer class.Alibaba’s financial affiliate (隶属机构), Huabei, makes small loans to online shoppers and vendors (小贩) on its e-commerce platforms. Peer-to-peer platforms have proved hugely popular in China, but the industry’s reputation has been ruined by scandal, like the (38)______ of Ezubao (易租宝).After that, regulators have stepped up their supervision of online lenders, which includes setting (39)______ on the amounts that can be borrowed. The regulators’ tightening management is“something like a shepherd herding (放牧) his sheep into a(n) (40)______ narrowing pen (围栏),” said Mark Natkin, the founder and managing director at Marbridge Consulting, based in Beijing.Ⅲ. Reading ComprehensionSection ADirections:For each blank in the following passage there are four words or phrases marked A, B, C and D. Fill in each blank with the word or phrase that best fits the context.“It will be marvelous. It will be spiritual.” This is the41 tagline that Virgin Galactic used in 2004 to attract potential customers to its planned space tourism services. It promised that, within five years, it would take atotal of over 3,000 passengers on life-changing trips in its spaceships. On July 11, 2021, after a last 90-minute delay, Virgin Galactic finally began its fulfillment of that original 42 . For four minutes, its six temporarily weightless passengers, including the firm’s British co-founder, Sir Richard Branson, saw the planet against the blackness of outer space.Back on the ground, Sir Richard called the experience “magical”. He may have43 the fact that he was able to reach space earlier than Jeff Bezos, a fellow billionaire but much wealthier. On July 20, Bezos would go slightly higher, for slightly less time, in a vehicle that had been built by his own spacefaring company, Blue Origin.The two tycoons (大亨) are among a growing number of 44 who believe that space tourism’s time has come. Suborbital (亚轨道的) tourism is part of a broader space economy that has rapidly grown over the past decade 45 technological advances. However, it is highly 46 that this will be true.For now, Blue Origin and Virgin Galactic will offer 47 suborbital flights to paying customers. Blue Origin is focusing much more on the development of a large new rocket that will be used for the launching of satellites, on selling advanced rocket engines to other companies, and on bidding (投标) on NASA contracts such as the recently announced plan to send humans back to the moon. 48 , Bezos doesn’t see Blue Origi n as a provider of services to adventure-seekers.Even this 4-minute suborbital travels have a(n) 49 , apparently. A major survey found that nearly two in five people with a net worth of over five million dollars would consider paying $250,000, Virgin Galactic’s current price, for a ticket. The business could be 50 , once regular flights begin to offset (抵消)the rockets’ development costs. But how fast and by how much is 51 . Without tourist-friendly destinations to visit (the capacity of the International Space Station is strictly limited), orbital tourism, with its far higher ticket prices, will not be a huge earner.Another challenge —and the industry’s biggest remaining uncertainty —relates to 52 . History has shown that a disaster, 53 in the early stages of an industry, can set progress back by years. NASA 54 its plan to send the untrained to orbit in 1986 after a school teacher was killed along with the rest of the crew in the Challenger tragedy. It was another 15 years before the next untrained person would 55 the journey on a Russian craft.41. A. modest B. demanding C. motivational D. unclear42. A. obligation B. commitment C. requirement D. survey43. A. felt ashamed of B. kept an eye on C. felt content with D. got upset with44. A. optimists B. opponents C. experts D. objectors45. A. thanks to B. but for C. apart from D. in spite of46. A. impossible B. undoubted C. probable D. favorable47. A. long-lasting B. inexpensive C. automatic D. brief48. A. On the contrary B. What’s more C. In the long run D. Above all49. A. shortcoming B. limitation C. advantage D. appeal50. A. profitable B. uneconomic C. eco-friendly D. globalized51. A. predictable B. essential C. irrelevant D. uncertain52. A. development B. security C. stability D. novelty53. A. particularly B. critically C. precisely D. unforeseeably54. A. initiated B. drafted C. modified D. suspended55. A. brave B. abandon C. steer D. pauseSection BDirections:Read the following three passages. Each passage is by several questions or unfinished statements. For each of them there are four choices marked A, B, C and D. Choose the one that fits best according to the information given in the passage you have just read.(A)About a quarter of the world drives on the left, and the countries that do so are mostly old British colonies like Australia, and Ireland. But Thailand, Indonesia and Japan have also developed this habit.This strange quirk (奇事) puzzles the rest of the world; however, there is a perfectly good reason. Up to the late 1700s, everybody travelled on the left side of the road because it is the sensible option for violent, feudal (封建的) societies of mostly right-handed people. Soldiers with their swords under their right arms naturally passed on each other’s right, and if you passed a stranger on the road, you walked on the left to ensure that your protective sword arm was between you and him.Revolutionary France, however, did away with this practice as part of its sweeping social rethink, and thanks to Napoleon, this change was carried out all over continental Europe. Because he was left-handed, his armies had to march on the right so he could keep his sword arm between him and any opponent. From then on, any country colonized by the French took to driving on the right.After the American Revolutionary War (1775-1783), the US became independent and decided to make traffic drive on the right in order to cut all remaining links to its British colonial past. Once America became the center of the car industry, if you wanted a good reliable vehicle, you bought American right-hand-drive. From then on, many countries changed out of necessity.Today, the EU would like Britain to fall into line with the rest of Europe, but this is no longer possible. It would cost billions of pounds to change everything round. The last European country to convert to driving on the right was Sweden in 1967. This ironically caused a reduction in car accidents because everyone drove carefully while getting used to the new system.56. Why did people travel on the left before the late 18th century?A. They were required to do so.B. They were mostly left-landed.C. It was easier to cross the street.D. They could feel safer from attacks.57. What was Napoleon’s attitude about walking on the left?A. Support.B. Disapproval.C. Doubt.D. Indifference.58. For Americans, driving on the right was a way to show ______.A. the connection with France was brokenB. the US was no longer ruled by the UKC. the American Revolution War had endedD. America was the center of the car industry59. What is the main reason for UK’s resistance to the European system?A. It would cost too much to change.B. Its increasing traffic accidents.C. Its influence on colonies.D. Its fast-developing car industry.(B)Amaze Y ourself, Amaze the WorldA job at Apple is unlike any other you’ve had. You’ll be challenged. You’ll be inspired. And you’ll be proud. Because whatever your job is here, you’ll be part of something big.Every detail matters.Every piece of packaging. Every swipe of the finger. Every “How can I help you?” Everything.And it doesn’t matter just some of the time. It matters all of the time. That’s how we do things at Apple. The result is some of the best-loved products in the world.Simplicity isn’t simple.Ask anyone here. It’s hard work. It means forever asking, “Why is it this way?” and “How can it be better?” It means rethinking every customer experience until the clutter has fallen away — until all that remains is what’s essential, useful and beautiful. That might be a new product feature that delights even die-hard fans. It might be a customer support call, or even a display in an Apple store, arranged and lit exactly so.Creativity from every corner.When you imagine the creative process at Apple, at first you may not picture someone in HR, or operations, or finance. But we expect creative thinking and solutions from everyone here, no matter what their responsibilities are. Innovation takes many forms, and our people seem to find new ones every day.60. Where will you probably find this passage?A. In a booklet introducing Apple’s management.B. On the official website of Apple Inc.C. In the help-wanted ads section of a newspaper.D. In a magazine telling business stories.61. The word “clutter” probably means ______.A. customers’ complaintB. simple questionC. unnecessary partD. old feature62. According to the passage, employees are expected to pay attention to the following EXCEPT ______.A. the wrapping of apple’s ne w productsB. a customer support callC. the accomplishments of AppleD. the lighting of an Apple store(C)The tendency to look for some outside group to blame for our misfortunes is certainly common and it is often strengthened by social prejudice. There seems to be little doubt that one of the principal causes of prejudice is fear —in particular the fear that the interests of our own group are going to be endangered by the actions of another. This is less likely to be the case in a stable, relatively unchanging society in which the members of different social and occupational groups know what to expect of each other and know what to expect for themselves. In times of rapid racial and economic change, however, new occupations and new social roles appear, and people start looking jealously at each other to see whether their own group is being left behind.When a community begins to feel unsure of its future, it becomes especially likely to turn in upon itself, to imagine that surrounding groups are threatening and unfriendly. At a time like this, distorted (扭曲的) ideas about theother community are readily believed and are passed on as statements of fact. One of the tragic things about intercommunal (社区间的)conflict is that both parties quickly find themselves “moral rationalization (合理化解释).” Each side insists and believes that its own actions are inspired by noble ideals, even when they are really acting out of pure self-interest. To a third party, neutral to the conflict, it may seem obvious that both are behaving unreasonably; but when one’s emotions are involved, and especially the emotion of fear, it is extremely difficult to remain sensible.Once prejudice develops, it is hard to stop, because there are often social forces at work which actively encourage unfounded attitudes of unfriendliness and fear towards other groups. One such force is education: We all know that children can be taught history in such a way as to keep alive old hatred and old prejudices between racial and political groups. Another social influence that has to be taken seriously is the pressure of public opinion. People often think and act differently in groups from the way they would do as individuals. It takes a considerable effort of will, and often calls for great courage, to stand o ut against one’s fellows and insist that they are wrong.Why is it that we hear so much more about the failures of relationships between communities than we do about the successes? I am afraid it is partly due to the increase in communication which radio, television and the popular press have brought about. In those countries where the media of mass communication are commercial enterprises, they tend to measure success by the size of their audience; and people are more likely to buy a newspaper, for instance, if their attention is caught by something dramatic, or something that can arouse their anxiety or interest as much as failures of intercommunal relationships do. And popular public speakers, especially if they are politicians addressing a relatively innocent audience, know that the best way to arouse such an audience is to frighten them.Where there is a real or imaginary threat to economic security, this is especially likely to arouse group prejudice. It is important to remember economic factors if we wish to lessen prejudice between groups, because unless they are dealt with directly, it will be little use simply advising people not to be prejudiced against other groups whom they see as their competitors, if not their enemies.63. People in a society of rapid racial and economic change ______.A. blame others for their recent misfortunesB. fear the loss of interests endangered by prejudiceC. are at a loss as to what to expect of others and themselvesD. are jealous of the new occupations and social roles64. Which of the following statements is NOT true about a community feeling unsure of its future?A. Distorted ideas about the other community are considered as facts.B. It’s extremely difficult for a third party to remain sensible and fearless.C. Old hatred and prejudices are passed on through history education.D. The pressure of public opinion makes people think and act in groups.65. The failures of relationships between communities are ______.A. commercial products of the media of mass communicationB. considered dramatic and arousingC. lessened due to real or imaginary threat to economic securityD. dealt with directly by advising people not to be prejudiced66. The passage is mainly about ______.A. how to eliminate social prejudiceB. what a sensible society is likeC. what causes social prejudiceD. how to build a sensible societySection CDirections:Read the following passage. Fill in each blank with a proper sentence given in the box. Each sentence can be used only once. Note that there are two more sentences than you need.A. She believes that grief and joy can coexist.B. She believes that joy is nowhere to be found.C. Heart-broken, her mother decided to try to give meaning to her grief.D. The training was really a way of going even deeper into others’ grief and joy.E. Rather, they support clients emotionally, physically, spiritually and practically.F. But the longer she took part in the program, the more she realized that she was only an expert in her grief.As parents of a child with a progressive and potentially fatal illness, Maryanne and Nick O’Hara lived on hope. Hope that their daughter, Caitlin, who was diagnosed with cystic fibrosis at age 2, would receive the lung transplant she spent two and a half years waiting for in her early 30s. Hope that her body wouldn’t reject it. That hope faded on Dec. 20, 2016, when Caitlin died of a brain bleed, two days post-transplant. She was 33.67 And so she signed for a certificate program to become an end-of-life doula (陪护), or “death doula”, working with individuals and families as they moved from this life into whatever is next.Unlike hospital workers, doulas don’t get involved in medical issu es. 68 And they step in whenever needed. That could be a few days before someone dies, sitting with them in their last hours, giving hand massages, chatting with them or even making snacks. Or it could be months or even years earlier, after someone receives a terminal diagnosis. They keep them company, listen to their life stories, or plan funerals.As for Ms. O’Hara, who is also a novelist, she is primarily helping people write their life stories.“I went into it thinking ‘I’ve been a volunteer with people who are dying. I’ve lost my loved one. I’m an expert in grief,’” she said.69“You really can’t tell anyone else how to feel grief,” Ms. O’Hara said. “You can offer advice, but there is no timeline for grief. As soon as people get a diagnosis, they are grieving. Their way of life is over.”70 “My grief is never going to go away,” she said. “I wouldn’t want it to. Grief and joy and love —it’s part of the same spectrum (图谱). I’m grieving because I loved someone so much.”Ⅳ. Summary Writing (10%)Directions:Read the following passage. Summarize the main idea and the main point(s) of the passage in no more than 60 words. Use your own words as far as possible.71. Coco-Cola’s Innovative Solution to Plastic PollutionPlastic pollution has become one of the most pressing problems of the 21st century with plastic waste almost everywhere. It is estimated that just in 2016, the world produced over 320 million tons of plastic, and this number is set to double by 2034.Plastic waste even makes its way into our oceans with 8 million pieces of plastic pollution per day polluting our waters and destroying the habitat for hundreds of species. It is now believed that there may be around 5.25 trillion pieces of plastic in the ocean weighing in total up to 269,000 tons.It’s been crystal clear for a long time that the situation is horrible, and something needs to be done, and fast.Luckily, Coca-Cola stepped up to the challenge and came up with one innovative and rather effective solution. The firm is replacing its plastic wrapping in Europe with a new paper board technology.This new tech is called the KeelClip and it is a first for the non-alcohol ready-to-drink (NARTD) industry. Using KeelClips, Coca-Cola aims to remove all plastic wrap from its can multipacks in all European Union markets by the end of 2021. The firm is optimistic that the move will save 2,000 tons of plastic and 3,000 tons of CO2 annually.“Innovation is a key principle of our sustainable packaging work and the application of this fully recyclable paperboard KeelClip™, which is composed of a top board that the cans clip into and a central cardboard ‘keel’ —similar to a ship’s keel — that stabilizes the pack, is another example of how we are delivering on our commitment to remove all unnecessary and hard-to-recycle single-use plastic from our products. We hope this eco-friendly initiative is a conscious approach to dealing with plastic pollution,” said Joe Franses, Vice President, Sustainability at Coca-Cola European Partners.Ⅴ. Translation (15%)Directions:Translate the following sentences into English, using the words given in the brackets.72. 随着经济的增长,人们对于奢侈品的需求也随之增长。

华为HCIP222认证考试题库(含答案)

华为HCIP222认证考试题库(含答案)

华为HCIP222认证考试题库(含答案)单选题(总共253题)1.运行MPLS设备的标签转发表中,对于不同的路由(但下一跳相同),出标签?A、一定不同B、一定相同C、可能相同答案:A解析:不同的路由,虽然下一跳相同,但是相当于不同的标签隧道,出标签一定不同。

所以正确答案是“一定不同”。

2.丢包仅发生在报文的发送端。

A、正确B、错误答案:B解析:丢包可能发生在数据转发任何环境。

所以题目中的描述是错误的。

3.如果防火墙的两个接口被划分到同一个区域,那么两个接口间的数据包流动也必须经过域间包过滤处理过程。

正确B、错误答案:A解析:防火墙上的不同接口之间传递数据时也会经过包过滤处理的过程。

所以题目中的描述是正确的。

4.在Peer-to-PeerVPN网络中,与客户端设备直接相连,并负责VPN业务接入运营商网络的设备是?A、CE设备B、PE设备C、P设备D、Client设备答案:B解析:与客户端CE直接相连的运营商端的设备成为PE设备,此设备负责VPN业务接入运营商网络。

所以正确答案是“PE设备”。

5.在下列的队列调度技术中,哪种队列调度技术的公平性较好?A、WFQPQC、FIFOD、RR答案:A解析:FIFO实现简单,处理速度快;RR避免了低优先级队列的“饿死”现象,不平等调度;PQ低延迟业务能得到保障;WFQ按权重实现公平调度;自动分类,配置简单。

所以正确答案是“WFQ”。

6.NFV中的VIM管理模块的主要功能包括资源发现、资源分配、资源管理以及? A、资源调度B、资源监控C、资源回收D、故障处理答案:D解析:VIM主要功能包括:资源的发现、虚拟资源的管理分配、故障处理等。

所以正确答案是“故障处理”。

7.关于802.1X和RADIUS两种技术的关系,下列选项描述正确的是?802·1X和RADLUS是同一组技术的不同名称B、802·1X是一个技术体系,它包含了RADIUS技术C、RADIUS是一个技术体系,它包含了802·1X技术D、802·1X和RADIUS是不同的技术,但经常配合在一起使用,共同完成对终端用户的准入控制答案:D解析:802.1X和RADIUS是不同的技术,通常radius作为服务器端进行验证,802. 1X作为准入控制。

对等网络课件108页PPT

对等网络课件108页PPT
用户需要某个音乐文件时,先 查询中央索引服务器,得到存 有该文件的节点信息。
用户选择合适的节点建立直接 连接。
Napster首先实现了文件查询 与文件传输的分离。
Napster的拓扑结构
2.2 全分布式非结构化拓扑
也称纯P2P结构,取消了中央服务器,每台机器是 真正的对等关系(称为对等机)。
自1999年以来,P2P研究得到学术界和商业组织的 广泛关注,同时该技术也一直饱受争议。
P2P技术被广泛应用于计算机网络的各个应用领域, 如文件共享、流媒体直播与点播、分布式科学计算、 语音通信、在线游戏支撑平台等。
目前以文件共享为代表的P2P应用已成为因特网上增 长最迅速的应用。
P2P技术的应用(1)
最早出现的P2P网络结构,也称集中目录式 结构,或非纯粹的P2P结构。
优:
维护简单,资源发现效率高。
缺点:
单点故障;扩放性差;版权问题。
对小型网络而言在管理和控制方面有一定优 势,不适合大型网络应用。
Napster文件共享系统
中央索引服务器保存所有用户 上传的音乐文件索引和存放位 置。
每个节点通过散列其IP地址, 得到一个128位的节点标识 符。
2. P2P网络的拓扑结构
P2P系统的主要概念之一是分散,包括分布式存储、 处理、信息共享和控制信息。
根据P2P系统的分散程度,可以将P2P架构分成纯分 散式和混合式。
根据结构关系可以将P2P系统细分为四种拓扑形式:
中心化拓扑 全分布式非结构化拓扑 全分布式结构化拓扑 半分布式拓扑
2.1 中心化拓扑
P2P技术的应用(3)
基于P2P方式的协同处理与服务共享平台:
P2P技术将众多终端空闲的CPU资源联合起来,服 务于一个共同的计算。

PEER-TO-PEER对等网络

PEER-TO-PEER对等网络
迟机制
根据网络中共享资源方式的不同,局域网有两种组织形式:一种是对等网络,另一种是工作站/服务器结构。对等网络中的计算机之间可以互相通信和共享资源(文件、外设等);而工作站/服务器结构的网络中可共享资源放在一台专用计算机-服务器(server)上,工作站之间不互相直接共享资源。对等网络比较灵活,适用于工作组级的小型网络,当网络规模较大时,其管理和安全性都变得比较困难,此时宜采用工作站/服务器结构。这两种结构一般是由系统软件的功能决定的。如用Windows NT Workstation就可以构成对等网络,专门的服务器则要使用Windows NT Server(或Novell Netware等)。
PEER-TO-PEER对等网络。

美国P2P网贷平台的法律规范及对中国的启示_以美国Prosper网站为例

美国P2P网贷平台的法律规范及对中国的启示_以美国Prosper网站为例

MoneyChina财经界美国P2P网贷平台的法律规范及对中国的启示———以美国Prosper网站为例中国人民银行张宏一、引言P2P (peer to peer )网络贷款基于互联网技术,脱离对银行等金融中介的依赖,由个人与个人之间直接进行资金借贷。

从2005年该业务诞生以来一直呈爆发性增长,截止2012年底,美国Prosper 网站会员人数超过100万名,累积发放借款4.43亿美元;英国Zopa 网站累积发放借款2.9亿英镑。

2012年当年,中国前十家网贷平台共计发放贷款94.9亿元,借款人共计9968人,放款人30634人。

有学者预测,由于网络交易速度快、交易成本低,基于互联网的金融新模式将在今后的20年中成为金融运行系统的主流。

在网贷平台的监管方面,美国著名网贷平台Prosper 已于2009年开始接受美国证监会的监管。

然而,我国网贷平台业务基本上处于“零监管”的状态,不时发生平台倒闭、平台负责人“跑路”、客户资金被网络黑客转移等重大风险事件。

从早期的“哈哈贷”、“淘金贷”到最近的“优易贷”,出问题的网贷平台涉及客户资金已超过千亿元。

为规范网贷业务发展、保障参与各方权益,亟需建立相关的法律规范和监管机制。

本文以美国的网贷平台Prosper 为例,简要分析美国网贷平台的法律框架和监管特点,得出对我国网络贷款平台监管规范的启示,并提出相关意见和建议。

二、美国P2P 网贷平台的基本法律框架(一)产品定义和业务流程美国的网络平台贷款定义“证券化产品”,平台的放款人在法律上被界定为购买Prosper 证券的投资者。

业务的基本流程如下:借款人在Prosper 平台发布借款信息,放款人对项目参与竞拍,同时,Prosper 对借款申请进行审核。

如果一个借款项目有足够多的竞价人且通过了Prosper 的审核,则该笔借款可以达成。

但借款资金并非由放款人直接借给借款人,而是由Ebank (一家注册于犹他州的银行)来发放,整个放款和贷后管理都由Ebank 来具体负责。

3GPP 5G基站(BS)R16版本一致性测试英文原版(3GPP TS 38.141-1)

3GPP 5G基站(BS)R16版本一致性测试英文原版(3GPP TS 38.141-1)

4.2.2
BS type 1-H.................................................................................................................................................. 26
4.3
Base station classes............................................................................................................................................27
1 Scope.......................................................................................................................................................13
All rights reserved. UMTS™ is a Trade Mark of ETSI registered for the benefit of its members 3GPP™ is a Trade Mark of ETSI registered for the benefit of its Members and of the 3GPP Organizational Partners LTE™ is a Trade Mark of ETSI registered for the benefit of its Members and of the 3GPP Organizational Partners GSM® and the GSM logo are registered and owned by the GSM Association

计算机三级网络 名词解释

计算机三级网络 名词解释

第一章1.对等网(peer to peer):如果每台计算机从逻辑上都是平等的,我们称作对等网。

2.客户/服务网(Client/Server):如果存在主从关系,我们称作非对等网络,又叫客户/服务网。

为主的计算机称为服务器,其余称为客户机。

3.ENIAC:世界上第一台计算机。

4.ARPANET:因特网(Internet)的前身。

5.TCP/IP:传输控制协议/网际协议,因特网的主要协议。

6.ISO:国际标准化组织。

7.CCITT:国际电报电话咨询委员会,现已取消,是ITU-T的前身。

8.Workstation:工作站。

9.MIPS:每秒百万条定点指令。

10.MFLOPS:每秒百万条浮点数指令。

11.bps:位每秒,数据传输率的衡量指标。

12.MTBF:Mean Time Before Failure 平均无故障时间,衡量可靠性的指标之一。

13.MTTR:Mean Time To Repair 平均故障修复时间,衡量可靠性的指标之一。

14.PCI:Peripheral Component Interface 外围部件接口,常用的局部总线。

15.SSE:Streaming SIMD Extensions流式的单指令多数据扩展,多媒体指令集。

16.EPIC:Explicit Parallel Instruction Computing简明并行指令计算,安腾的主要特点。

17.VESA:Video Electronic Standard Association 视频电子标准协会制定的局部总线标准。

18.Inte:因特尔,全球最大的中央处理器厂商和著名的网络产品厂商。

19.Microsoft:微软,全球最大的软件厂商。

20.IBM:美国商用机器,全球著名的IT厂商,以高技术著称。

21.SUN:斯坦福大学网络公司,全球著名的网络产品软硬件厂商。

22.JPEG:由ISO和CCITT联合制定的静态图像压缩标准。

23.MPEG:由ISO制定的多媒体视音频压缩标准。

p2p通信原理

p2p通信原理

p2p通信原理P2P通信原理P2P(Peer-to-Peer)通信是指在计算机网络中,两个或多个计算机之间直接进行通信,而不需要经过中心服务器的中转。

P2P通信原理基于对等网络结构,每个参与者都可以作为客户端和服务器同时存在,相互之间可以直接交换数据。

P2P通信的出现,使得信息传输更加高效、灵活,并且可以更好地应对服务器故障或网络拥堵的情况。

P2P通信的基本原理是通过建立直接的点对点连接来实现。

在传统的客户端-服务器模型中,客户端向服务器发送请求,服务器对请求进行处理并返回相应的数据。

而在P2P通信中,每个节点都可以是客户端和服务器,节点之间可以直接发送和接收数据。

这种对等网络结构使得数据传输更为高效,减少了单一服务器的负载压力,并且可以更好地应对服务器故障的情况。

P2P通信的实现依赖于一些关键技术。

首先,P2P通信需要节点之间建立直接的连接。

为了实现这一点,需要使用一些协议和技术,如NAT穿透、STUN和TURN等。

NAT穿透技术可以绕过网络地址转换(NAT)设备,使得节点之间可以直接通信。

STUN和TURN技术则可以帮助节点发现对方的真实地址和建立中转连接。

P2P通信需要一种方式来管理节点之间的连接。

通常使用的方式是通过一个中心服务器来协调节点之间的连接,这个服务器被称为超级节点或种子节点。

超级节点负责帮助节点发现其他节点,并协调节点之间的连接。

一旦节点之间建立起连接,它们就可以直接交换数据,不再依赖于超级节点。

P2P通信需要一种方式来管理节点之间的数据传输。

通常使用的方式是基于流式传输的协议,如TCP或UDP。

TCP协议提供可靠的数据传输,保证数据的完整性和有序性,适用于对数据传输质量要求较高的场景。

而UDP协议则提供了更低的延迟和更高的传输速率,适用于实时性要求较高的场景。

总结一下,P2P通信原理是基于对等网络结构的,每个节点都可以作为客户端和服务器,节点之间可以直接发送和接收数据。

P2P通信的实现依赖于一些关键技术,如NAT穿透、STUN和TURN等,用于建立直接的连接;超级节点用于协调节点之间的连接;基于流式传输的协议用于管理节点之间的数据传输。

美军全球信息栅格(GIG)简介

美军全球信息栅格(GIG)简介
24
IPv6的优点
❖ 地址空间大 ❖ 提供端到端的安全性能 ❖ 有较高的服务质量保证 ❖ 移动计算功能强 ❖ 网络运营管理简便
25
向IPv6过渡是GIG的必然选择
❖ 为体现“网络中心战”的思想,提高系统的安全性能, 使不断增长的传感器、武器平台、信息设施和各类人 员互联起来,采用IPv6将是一个明智选择。
❖ 美军充分认识到这一决策的挑战性和艰巨性,强调要采 用成熟的民用技术和产品(COTS),分步实施。
❖ 过渡期间,采用试点(pilot implementation)、测试床 (test bed)、新技术演示(demonstration)等技术手段。
❖ 美国防部已成立了IPv6过渡办公室,指定国防信息系统 局负责IPv6地址的分配、注册和控制,并从规划计划、体 系结构和设计开发标准和产品等方面为过渡奠定基础。
36
今年春天,美国国家科学资金委员会资助的 最大科研网 TeraGrid 中的超级计算机(位于 伊利诺斯州立大学的超级计算机应用中心 NCSA和圣迭戈大学的超级计算中心SDSC)遭 遇到黑客的持续攻击,连续数天处于瘫痪状态, 这是一个主要应用于天气预报和基因排序的大 规模数据处理的科研网格。
37
GIG的信息保障
❖ 不同领域的用户对Grid的认识有很大差别, 切入的角度和研究的重点有所不同,因而赋 予Grid不尽相同的体系结构和使命任务。
❖ 学术界比较看重 Grid 的规模和合作求解问题 的能力,他们最早提出“网格协议体系结构” GPA,又称“沙漏模型” 。
7
网格协议体系结构(GPA)
工具与应用
目录代理 诊断与监控等
平台,统一资源调度、访问视图和服务管理。
5
2002年7月,Ian Foter提出判断 Grid 的三项标准:

基于Gnutella协议的Peer-to-Peer网络连接管理

基于Gnutella协议的Peer-to-Peer网络连接管理

摘要Peer-to-peer(P2P)是通过直接交换共享计算机资源和服务的一种网络体系结构。

Gnutella作为一种典型的P2P网络通信协议,具有智能发现节点、完全分布式的特点,它能够有效的消除单点瓶颈,使网络具有很强的健壮性。

与此同时,该协议也产生了以指数级增长的冗余消息,效率低下,使之只能在小规模的网络上使用,很难成为应用的主流。

本文从Gnutella协议的节点定位入手,提供了两种节点定位模型:网络模型和多播模型。

这两种模型都基于Gnutella网络完全分布式和动态性的特点,能够实现节点之间的快速定位,从而获取有效的节点连接信息。

我们详细解析了Gnute|la网络的广播机制,划分了Gnutella协议中四种消息的优先级,分析了丢弃连接的必要性和可行性。

在此基础上,我们提出了F-Measure算法。

同时,考虑到Gnutella协议其下层的连接大多是基于TCP和PPP协议的,因此,本文还借鉴了当前Intemet上广泛使用的数据流控制方法——zigzag算法,并对其进行了改进,使其能够更好的辅助F-Measure算法进行Gnutella网络的连接管理。

实验数据表明:我们的F-Measure算法及其辅助管理手段能够极大地减少消息的冗余量和极好地适应网络流量的动态变化。

关键词:P2P,Gnutella,网络模型,多播模型,F-Measure算法,缓冲时间Zig—Zag算法AbstractAbstractPeer—.to--Peeristhearchitectureofnetworkbydirectlyexchangingandsharingresou,rcesandserviceofcomputer.GnutellaasatypicalP2Pnetworkcommunicationprotocol,canintelligentlyfindnodes.Moreoverithasentirelydistributedcharacter.ItCalleffectivelyavoidisolationnodesbottleneckandmakenetworkmuchmorerobust,Butatthesarfletime,theGnutellaprotocoIalsoproducesexponentiallyincreasingredundancemessageanditsefficiencyisverylow.Itonlyapplythelow-scalenetworkanditisverydifficulttorumtomainstreamapplication.ThePaDerstartsfromtheorientationamongstthecrunodesoftheGnutellaprotocolandputforwardtotwomodelsoftheorientation.Theyarenetworkmodelandmaltibroadcastmodel.Thetwomodelsbothbaseontheentirelydistributedanddynamiccharacter.Theycanfindtheneighbournodesquicklyandgettheconnectioninformationofthenode.111ePaDerresolvesthenetworkmessage.broadcastedmechanismofGnutellaindetail,dividespriorlevelofthefourkindsofmessageandanalysethenecessityandfeasibilityofdiscardingtheconnection.WebringsforwardF-Measurealgorithm.Atthesametime,takingtheTCPandPPPintoaccountwhicharethebasesoftheunderlayerofGnutellaprotocol,weusethethoughtofdataflowforreferencewhichiswidelyappliedintheIntemet.WemodifytheZig-ZagalgorithminordertohelpF—Measurealgorithmmanagenetworkconnection.TheexperimentdatashowsthatourF-Measurealgorithmandassistantmanagingmeanscanhugelydecreasetheamountofrednndancemessageandbetterfittlledynamicchangeofnetworkflux.KeyWords:P2P,Gnutella,networkmodel,multi—broadcastmodel.F—Measurealgorithm,buffer-time,Zig-ZagalgorithmII基于Gnutella协议的Peer.to-Peer网络连接管理序吾P2P(Peer.to.Peer)是通过直接交换共享计算机资源和服务的--}ee网络体系结构,它引导网络应用的核心从中央服务器向网络边缘的终端设备扩散。

P2P流量识别技术综述

P2P流量识别技术综述

第29卷2009年12月计算机应用JournalofComputerApplicationsV01.29Dee.2009文章编号:1001—9081(2009)s2—0011—05P2P流量识别技术综述袁雪美,王晖,张鑫,刘亚杰(国防科学技术大学信息系统与管理学院,长沙410073)(xuemeiqiqi.yuan@gmail.咖)摘要:在归纳P2P流量识另q问题概念的基础上,对现有的P2P流量识别技术进行了较全面地分析。

借助分类模型形式化地定义P2P流量i7,Aq问题,依据所采用的识别特征将已有技术分为基于端口号、基于流量特征、基于应用层签名、基于双重特征和基于统计行为特征五类方法,并对各类方法进行了介绍、分析与优劣对比。

探讨了新兴的P2P流媒体流量识别问题,总结了P2P流量识别技术的发展趋势。

关键词:P2P流量识别;端口号;流量特征;应用层签名;混合特征;统计行为特征中图分类号:TP393.07文献标志码:ASurveyofP2PtraffiCidentificationYUANXue—mei,W.ANGHui,ZHANGXin,UUYa-jie(Collegeo,Information跏觚andManagement,NationalUnivers蚵of蜘eTechnology,ChangshaHunan410073,China)Abstract:AcomprehensivestaayofexistedP2Ptramcidentificati∞methodbasedonthedefinitionofP2PtrafficidentifieationWaSintrodueed.n地problemofP2Ptra缶cidentificationwasdefinedthoughtheclassificationmodel.Fivemethodswhichwereport—based,trafficfeaturesbased。

Peer-to-peer presence detection

Peer-to-peer presence detection

专利名称:Peer-to-peer presence detection发明人:Mohamed M. Abdelaziz,Eric Pouyoul,Jean-Christophe Hugly,Bernard A.Traversat,Michael J. Duigou申请号:US10263923申请日:20021003公开号:US20030041141A1公开日:20030227专利内容由知识产权出版社提供专利附图:摘要:Embodiments of a decentralized mechanism for detecting the presence ofentities in a peer-to-peer network. In one embodiment, using pipes, a peer may uniquelyand independently generate an identity for a peer-to-peer network entity. The identity may be used to probe the peer-to-peer network for presence of instances of the corresponding entity on other peers regardless of which peer the identity was generated on and without using a central identity repository or presence system. In one embodiment, net crawling may be used to map identities and resolve the locations of instances of the corresponding entities. An entity may move anywhere on the network, and the decentralized mechanism for detecting entity presence may be used to locate the instance(s) of the entity.申请人:ABDELAZIZ MOHAMED M.,POUYOUL ERIC,HUGLY JEAN-CHRISTOPHE,TRAVERSAT BERNARD A.,DUIGOU MICHAEL J.更多信息请下载全文后查看。

Peer-to-peer communication

Peer-to-peer communication
申请人:クゥアルコム・インコーポレイテッド 地址:アメリカ合衆国、カリフォルニア州 92121-1714、サン・ディエゴ、モアハウ ス・ドライブ 5775 国籍:US 代理人:蔵田 昌俊,河野 哲,中村 誠,福原 淑弘,峰 隆司,白根 俊郎,村松 貞男,野河 信久,幸長 保次郎, 河野 直樹,砂川 克,勝村 紘,河井 将次,佐藤 立志,岡田 貴志,堀内 美保子,竹内 将訓,市原 卓三,山下 元 更多信息请下载全文后查看
专利内容由知识产权出版社提供
专利名称:Peer-to-peer communication 发明人:マシュー・エス.・グロブ,ラィスター,クレ イン・エス.・ギルハウセン,ラミン・レザイー ファー 申请号:J P 2009274 973 申请日:20091202 公开号:J P 4 763828B 2 公开日:20110831
摘要:The invention relates to a method of operating an access terminal (106) for peer-to-peer communication in a multiple-access network (100), comprising: transmitting peer-to-peer communications (110) to at least one peer access terminal (106) on a reverse link frequency of the multiple-access network (100); and receiving peer-to-peer communications (110) from at least one peer access terminal (106) on the reverse link frequency of the multiple-access network (100).

区块链关键技术试题及答案

区块链关键技术试题及答案

区块链关键技术试题及答案1. [单选]以下对分布式存储描述正确的是:()A.分布式存储是将数据分散存储到多个服务器上B.分散的存储资源构成一个虚拟的存储设备C.数据分散地存储在服务器集群的各个角落D.以上都对(正确答案)2. [单选]区块链是一种“特殊”的分布式存储,哪个不是其特性:()A.节点无需同步数据(正确答案)B.数据是由系统所有参与者来集体维护C.没有中心节点,每个节点都是平等的(P2P)D.每个节点都保存数据副本3. [单选]中心化的分布式网络与其它P2P网络相比,哪个描述是不正确的:()A.两个peer之间的通信要通过中心服务器B.可靠性高(正确答案)C.可维护性高D.发现算法效率高4. [单选]以下对P2P(peer-to-peer)网络描述正确的是:()A.节点无主从之分B.耐攻击C.高容错D.以上都对(正确答案)5. [单选]以下选项中,哪些不是联盟链区块头中包含的信息?B.当前哈希C.前续哈希D.时间戳(正确答案)6.[单选]默克尔树是迭代哈希,每个区块包含多笔交易,交易()进行哈希运算,()迭代哈希,直到最后得到一个哈希值,最终生成的哈希值叫默克尔树根。

A.两两组合,从上向下B.两两组合,从下向上(正确答案)C.多个组合,从下向上D.多个组合,从上向下7. [单选]前续区块哈希是指()的哈希值,通过对()的()数据进行哈希计算得出。

A.前一区块前一区块区块头(正确答案)B.前一区块前一区块区块体C.后一区块后一区块区块头D.后一区块后一区块区块体8. [单选]区块链的去中心化特点,以下说法不正确的是?A.没有中心服务器B.所有节点权限对等C.数据分布存储D.系统低冗余(正确答案)9. [单选]交易数据是写入区块链的哪个部分:B.区块体(正确答案)C.默克尔树D.以上都不是10. [单选]区块链由哪个特点决定了其不可篡改性:()A.本区块的哈希值与本区块的数据内容相关联B.修改任何数据都会影响到本块哈希值C.下一区块会记录前续区块哈希值D.以上都是(正确答案)11. [单选]区块链中,前后区块通过哪种密码学技术形成指针使区块依次相连?()A.对称加密B.非对称加密C.哈希算法(正确答案)D.零知识证明12. [单选]以下描述对称加密算法的说法不正确的是:()A.国密SM4是对称加密算法B.发收信双方各自使用不同的密钥(正确答案)C.解密方事先必须知道加密密钥D.发收信双方都使用一个密钥13. [单选]比特币系统中,可以把地址理解为:()A.公钥(正确答案)B.私钥C.签名14. [单选]非对称加密的特点是:()A.安全性更高B.公钥公开,私钥仅自己保留C.加密和解密花费时间长D.以上都是(正确答案)15. [单选]CA中心的特点是:()A.具权威性B.依赖度极高的第三方C.其资格证书经国家颁发D.以上都是(正确答案)16. [单选]区块链主要运用数字证书技术的是:()A.公链B.比特币C.联盟链(正确答案)D.以上都是17. [多选]P2P(peer-to-peer)又称为:()A.对等式网络(正确答案)B.非对等式网络C.中心化网络D.点对点网络(正确答案)18. [多选]P2P 网络有那些分类:()A.中心化 P2P 网络(正确答案)B.全分布式非结构化 P2P 网络(正确答案)C.全分布式结构化 P2P 网络(正确答案)D.半分布式 P2P 网络(正确答案)19. [多选]分布式存储分类:()A.分布式文件系统(正确答案)B.分布式Key-Value系统(正确答案)C.分布式关系型数据库系统(正确答案)D.数据仓库20. [多选]关于比特币挖矿的说法正确的是:()A.挖矿相当于做数学题(正确答案)B.经过大量的计算(哈希碰撞),试出一个正确的区块哈希值(正确答案)C.矿工用来争夺记账权的手段(正确答案)D.是矿工的工作量证明(PoW)方式(正确答案)21. [多选]区块链的特性有哪些?()A.去中心化(正确答案)B.集体维护(正确答案)C.系统透明(正确答案)D.不可篡改(正确答案)22. [多选]区块链不是一项新技术,而是新的技术组合,其关键技术包括:()A.分布式(正确答案)B.共识机制(正确答案)C.密码学(正确答案)D.智能合约(正确答案)23. [多选]区块由什么组成?A.区块头(正确答案)B.前续哈希C.区块体(正确答案)D.默克尔树根24. [多选]区块链的哪些表现体现了其集体维护的特性:()A.每个节点只保留部分数据副本B.去中心化使所有节点都有数据副本(正确答案)C.每个区块都存有前续哈希(正确答案)D.前续区块哈希与其自身区块交易数据无关25. [多选]对称加密算法相比非对称加密算法:()A.对称加密算法加解密效率更高(正确答案)B.对称加密算法加解密速度更快(正确答案)C.对称加密算法密钥只有一个(正确答案)D.对称加密算法的密钥分发和管理比较困难(正确答案)26. [多选]哈希算法特性包含以下哪几种?()A.不可逆(正确答案)B.任意长度输入(正确答案)C.固定长度输出(正确答案)D.每个输入对应唯一的输出(正确答案)27. [多选]以下哪些是密码学中的常见术语?()A.密钥(正确答案)B.明文(正确答案)C.密文(正确答案)D.分布式网络28. [多选]哈希算法又叫做:()A.散列函数(正确答案)B.对称加密函数C.杂凑函数(正确答案)D.椭圆曲线函数29. [多选]以下哪些是哈希算法的应用场景:()A.对称加密B.数字签名(正确答案)C.防篡改验证(正确答案)D.保存用户密码(正确答案)30. [多选]公钥和私钥成对使用的描述正确的是:()A.公钥加密,公钥解密B.私钥签名,私钥验签C.私钥签名,公钥验签(正确答案)D.公钥加密,私钥解密(正确答案)31. [多选]下面有关数字签名的描述正确的是:()A.数字签名技术融合了非对称加密技术和数字摘要(哈希算法)技术(正确答案)B.数字签名是对发送者发送信息的有效证明(正确答案)C.接收者使用发送者的公钥才能查看信息(正确答案)D.通过数字签名可以验证消息发出者身份和消息是否被篡改(正确答案)32. [多选]下面有关数字证书的描述正确的是:()A.数字证书是由CA机构颁发的数字认证(正确答案)B.数字证书用来识别网络上的用户身份(正确答案)C.数字证书颁发、验证环节需要CA机构的私钥签名和对明文进行哈希运算(正确答案)D.联盟链中的CA节点负责数字认证(正确答案)33. [多选]数字证书的特性表现在:()A.用户可即时申请和开通(正确答案)B.用户不需要掌握加密技术(正确答案)C.数字证书由CA中心签发(正确答案)D.数字证书是唯一的(正确答案)34. [多选]零知识证明的特点是:()A.证明者不向验证者提供任何有用信息,使验证者相信某个论断是正确的(正确答案)B.在泄露极少量数据的情况下,使验证者相信某个论断是正确的C.零知识证明并不是数学意义上的证明(正确答案)D.零知识证明是概率证明而不是确定性证明(正确答案)35. [判断]非对称加密比对称加密更高效。

常用精准广告英文缩写

常用精准广告英文缩写

1.R T B R e a l T i m e B i d d i n g[r iəl t a i mˈbɪdɪŋ]实时竞价2.SSP Supply Side Platform [səˈplai said ˈplætfɔ:m ] 供应方平台3.Ad Exchange [æd iksˈtʃeindʒ] advertisement exchange广告交易平台4.DSP Demand-Side Platform [dɪˈmænd said ˈplætfɔ:m ] 需求方5.DMP Data-Management Platform [ˈdetəˈmænɪdʒmənt ˈplætfɔ:m ]数据管理平台6.Direct-Response Advertising [[diˈrekt riˈspɔns ˈædvəˌtaɪzɪŋ]直接反应广告]] 旅游目的地营销系统19.Ambush Marketing [ˈæmbʊʃˈmɑ:kɪtɪŋ] 伏击营销20.CPM Cost Per Mille [kɔst pə: ˈmili: ] 千人成本21.ESP Emotional Selling Proposition [iˈməuʃənl ˈseliŋˌprɔpəˈziʃən ] 情感销售主张22.Viral Marketing [ˈvaɪərəl ˈmɑ:kɪtɪŋ] 病毒性营销23.Mini-site [ˈmɪni sait ] 微型网站24.DAGMAR模式Defining Advertising Goals for Measured Advertising Results[diˈfainiŋˈædvəˌtaɪzɪŋɡəul z fɔ: ˈmeʒəd ˈædvəˌtaɪzɪŋriˈzʌlts]为度量结果而确定广告目标25.Ad words [æd wə:dz] 关键词广告26.Digital Marketing [ˈdidʒitəl ˈmɑ:kɪtɪŋ] 数字营销27.Differentiated Marketing [ˌdifəˈrenʃieitid ˈmɑ:kɪtɪŋ] 差异化营销28.Online Marketing [ˌɔnˈlainˈmɑ:kɪtɪŋ] 在线营销29.Pareto principle [pəretəu ˈprinsəpl]马特莱定律,又称“二八定律”ns]n]、42.B2F Business To Family [ˈbiznis tu: ˈfæmili]商业机构对家庭消费的营销商务43.CPS Commodity Promotion Solution [kəˈmɔditi prəˈməʊʃən səˈlu:ʃən]商品推广解决方案44.MU Marginal Utility [ˈmɑ:dʒinəl juˈtiliti] 边际效用45.PR on line Public Relations on line [ˈpʌblik riˈleiʃənz ɔn lain]网络公关46.Event Marketing [iˈventˈmɑ:kɪtɪŋ]事件营销47.Media Convergence [ˈmi:diəkənˈvə:dʒəns]媒介融合48.Customer to Business [ˈkʌstəmətu: ˈbiznis ] 消费者对企业49.e-business [i: ˈbiznis ]电子商务50.Marketing Plan [ˈmɑ:kɪtɪŋplæn]营销策划51.Web Data Mining [web ˈdeitəˈmaɪnɪŋ]数据挖掘52.GRP Gross Rating Points [ɡrəusˈreitiŋpɔints]毛评点53.Pay-Per-Call[pei pə: kɔ:l ]呼叫广告,又可称为按通话付费广告P Unique Selling Proposition [ju:ˈni:kˈseliŋ]独特的销售主张])71.Advertising Effect [ˈædvəˌtaɪzɪŋiˈfekt] 广告效应72.KPI Key Performance Indicator [ ki:pəˈfɔ:məns ˈɪndɪˌkeɪtə] 关键绩效指标73.TGI指数Target Group Index [ˈtɑ:ɡitɡru:pˈindeks ] 目标群体指数mission [kəˈmiʃən ] 佣金75.Advertisement Management [ədˈvə:tismənt ˈmænidʒmənt ] 广告管理76.SEO Searching Engine Optimization [ˌsɜ:tʃɪŋˈendʒin ˌɔptimaiˈzeiʃən ] 搜索引擎优化77.VTR View-Through-Rate [vju: θru:reit ] 显示到达率78.The Long Tail [ðəlɔŋteil ] 长尾理论79.On-line Marketing [ˈɔ:nˌlainˈmɑ:kɪtɪŋ ]或E-Marketing [ i: ˈmɑ:kɪtɪŋ] 网络营销80.CRM Customer Relationship Management [ˈkʌstəməriˈleiʃənʃip ˈmænidʒmənt ]客户关系管理81.Banner [ˈbænə] 旗帜广告82.Affiliate Marketing [əˈfilieit ˈmɑ:kɪtɪŋ] 加盟营销83.Dynamic Rotation [daiˈnæmik rəʊˈteɪʃən ] 动态轮换]98.BANNER [ˈbænə] 横幅广告、旗帜广告99.BUTTON [ˈbʌtən ] 按钮广告100.Pop-up ads [pɔp ʌp ædz ] 弹窗、弹出窗口广告101.Floating ads [ˈfləʊtɪŋædz ] 浮动广告102.Interstitial [ˌintə(:)ˈstiʃəl ] 插播式广告103.Full Screen [ful skri:n ] 全屏广告104.SNS Social Network Service [ˈsəuʃəl ˈnetwə:k ˈsɜːvɪs ] 社交网络服务Social Network Software [ˈsəuʃəl ˈnetwə:k ˈsɔftwɛə] 社交网络软件Social Network Site [ˈsəuʃəl ˈnetwə:k sait ] 社交网络站点105.Behavior Targeting [bɪˈheɪvjəˈtɑ:ɡitɪŋ] 行为定向匹配模式106.Content Targeting [ˈkɔntent ˈtɑ:ɡitɪŋ] 内文匹配模式107.Demographics Targeting [ˌdeməˈgræfɪks ˈtɑ:ɡitɪŋ] 人口数据匹配模式(代理商模式)108.SPIN Situation Question [ˌsitjuˈeiʃən ˈkwestʃən ] 情况问题状况询问Problem Question [ˈprɔbləm ˈkwestʃən ] 难点问题问题询问ɔv。

  1. 1、下载文档前请自行甄别文档内容的完整性,平台不提供额外的编辑、内容补充、找答案等附加服务。
  2. 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
  3. 3、如文档侵犯您的权益,请联系客服反馈,我们会尽快为您处理(人工客服工作时间:9:00-18:30)。

第27卷第2期2009年3月应用科学学报JOURNAL OF APPLIED SCIENCES—Electronics and Information EngineeringVol.27No.2Mar.2009文章编号:0255-8297(2009)02-0124-07基于贝叶斯网络的Peer-to-Peer识别方法李君1,2,张顺颐1,王浩云1,李翠莲21.南京邮电大学信息网络技术研究所,南京2100032.浙江万里学院通信工程系,浙江宁波315100摘要:网络业务分类与识别是网络管理、网络规划和安全的必要途径,而Peer-to-Peer(P2P)流量由于采用伪装端口、动态端口以及应用层加密,已成为业务分类与识别的主要难点.该文提出了P2P业务的精确识别方法,通过对流统计特性的分析,提取相关特征属性,应用遗传算法选取最优特征属性子集,并采用贝叶斯网络机器学习方法识别P2P流量.实验表明K2,TAN和BAN能有效快速地识别P2P业务,分类精度高达95%以上,很大程度上优于朴素贝叶斯分类和BP神经网络方法.同时该系统具有可扩展性,能够识别未知的P2P流量,并适用于实时分类识别环境.关键词:Peer-to-Peer;流量识别;朴素贝叶斯;贝叶斯网络中图分类号:O212.7文献标志码:APeer-to-Peer Traffic Identification Using Bayesian NetworksLI Jun1,2,ZHANG Shun-yi1,WANG Hao-yun1,LI Cui-lian21.Institute of Information Network Technology,Nanjing University of Posts and Telecommunications,Nanjing210003,China2.Department of Telecommunication Engineering,Zhejiang Wanli University,Ningbo315100,Zhejiang Province,ChinaAbstract:Accurate traffic classification is vital to numerous network activities,such as security monitoring,qual-ity of service provisioning and network planning.However,current P2P applications,which generate a substantial volume of Internet traffic,use dynamic port numbers,HTTP masquerading and payload encryption to avoid de-tection.In this paper,we present an accurate P2P identification method using Bayesian networks.Based on the abstracted attributes offlow statistics,the optimal attribute subset is selected using genetic algorithms and P2P traffic is identified using Bayesian networks.We evaluate the algorithms and compare them to the previously used Naive Bayesian model and BP perceptron.Experimental results show that the proposed algorithms achieve better overall accuracy up to95%with less cost.Further,our result indicates that the approaches are capable of identifying unknown P2P traffic and applicable to the real-time applications.Keywords:Peer-to-Peer,traffic identification,naive Bayes,Bayesian networksP2P网络因具有强大的内容发布、搜索和共享能力而得到广泛应用,P2P流量已逐步取代HTTP流量成为Internet流量的主体.迅速增长的P2P流量消耗了大量的网络带宽,甚至引起网络拥塞,极大地降低了网络性能,妨碍了正常网络业务的开展.此外,通过不安全的P2P网络环境获得的应用程序给网络安全带来了极大隐患,也为病毒传播和恶意攻击创造了条件.因此,P2P流量的分析与识别已成为网络监测与管理的收稿日期:2008-08-18;修订日期:2008-12-22基金项目:国家“863”高技术研究发展计划基金(No.2005AA121620,No.2006AA01Z232);浙江省自然基金(No.Y1080935);江苏省普通高校研究生创新计划基金(No.CX07B_110z)资助项目作者简介:李君,副教授,博士生,研究方向:计算机通信网络与IP技术、P2P技术、网络业务流量识别与分类、分布式网络管理,E-mail: lijunreed@第2期李君等:基于贝叶斯网络的Peer-to-Peer识别方法125热点.常用的P2P流量识别方法有端口映射法和净荷分析法.早期的P2P协议一般采用固定端口,因此可以直接根据端口号精确识别[1-2].但目前很多新出现的P2P应用采用随机端口,甚至伪装成其他熟知端口(如80端口)穿越防火墙,致使简单的端口识别法失效.净荷分析法是目前最精确的P2P识别方法[3-4],它根据已知的特征字符串匹配应用层净荷,所以不能识别未知或加密的P2P应用.而且随着特征字符串数量剧增,特征长度多变且位置不定,使得模式匹配的难度进一步增大,限制了净荷分析法的应用.为了克服上述识别方法的限制,基于流量特征分析的P2P业务识别方法[5-7]应运而生.它通过对传输层数据包(包括TCP和UDP数据包)的分析,结合P2P系统所表现出来的流量行为特征和流统计特性来区分P2P流量,并引入机器学习方法以处理大量特征数据,挖掘P2P流量的本质特征.Zander等[8]尝试基于最大似然的业务分类方法,Erman等[9]则采用非监督的聚类算法K-Means和DBSCAN来分类网络流量,并证明两种方法都比AutoClass算法速度快且分类效果好. Moore和Zuev[10]利用核估计朴素贝叶斯方法实现业务分类,他们还通过减少特征复杂度和训练集大小,研究基于Netflow流量的轻量级流量分类技术[11].然而,上述方法的分类精度并不理想.为了提高分类精度,Rui等[12]应用优化的支持向量机识别P2P流量,Auld等[13]引入了复杂的贝叶斯神经网络方法分类网络流量.两种方法都具有较高的分类精度和稳定性,但计算复杂度高,根本无法适用于在线流量分类.于是,根据业务分类与识别的研究实践与经验,本文选取相关的流统计特征,采用基于遗传算法的特征选择方法筛选出最优特征子集,最后应用贝叶斯网络学习方法实现P2P业务的识别.实验结果表明本文提出的方法优于文献[10]中所采用的朴素贝叶斯分类法,它不仅具有较高的分类精度,而且不受应用层净荷加密的限制,能够识别未知和新出现的P2P应用流量,并能适用于在线实时P2P流量识别.1基于机器学习的P2P流量识别方案基于机器学习(ML)的P2P流量识别方案由两部分组成:ML建模和ML分类,如图1所示.首先,系统抓包获取网络业务分组,按五元组(即源地址、目的地址、源端口、目的端口和传输层协议)分流,提取相应的分组头部信息,计算流统计特征值并置入流统计特性数据库.抽样选取一定的样本数,形成训练测试集.为了去除特征属性的冗余和相关,先执行特征选择,然后在得到的最优特征子集上选用合适的ML算法进行训练,最后输出ML分类模型.在ML测试/分类阶段,同样抓取网络流量,根据选定的最优特征属性子集统计流量特征值,由ML分类模型对测试集或该流量统计数据进行分类,其结果可用于各种网络应用,如QoS定制和计费等.由于网络新业务不断出现,网络流量复杂多样且动态变化,所以要定期地对ML分类模型进行离线更新.图1基于机器学习的P2P流量识别方案Figure1Flow chart of P2P traffic identificationbased on machine learning1.1基于遗传算法的特征属性选择业务流的统计特征反映了业务流的本质,用以区分不同的业务流.一般情况下,特征越多,描述的实例越充分,越能精确区分P2P流和非P2P流.但获得特征数据不仅需要大量的测量时间,而且样本存储占用的空间以及样本处理时的内存与CPU占有率也急剧增加,计算时间大大增加,无法适用于实时分类.本例中虽然已经根据协议分析和网络流量统计特性分析,从大量的属性特征(多达100多个[14])中筛选出20个左右最具代表性的特征,但仍然存在无用或冗余的属性,不仅增加了计算复杂度,还会影响学习质量,所以提取最能刻画样本数据的最小特征子集是非常重要的.本文采用遗传算法选择最优特征子集.1.1.1遗传算法基本原理遗传算法GA(genetic algorithm)是在生物进化论和遗传学的基础上产生和发展起来的一种全局启发式优化算法.GA以编码空间代替问题的参数空间,以适应度函数为评价依据,以编码为进化基础,以对群体中个体位串的遗传操作实现选择和遗传机制,建立起一个迭代过程.在这个过程中,通过选择、交叉和变异等操作重组编码位串中的重要基因(基因对应于本文所考虑的流统计特征),使新一代的位串集合优于老一代的位串集合,群体的个体不断进化,最终收敛于最优状态.1.1.2特征子集选择的GA方法本问题是关于两类分类问题(即P2P和非P2P)的特126应用科学学报第27卷征子集选择.在原始特征集合U n 空间下的实例空间X ={x 1,x 2,...,x m },x i =(x i 1,x i 2,...,x in ),实例总数为m ,特征数为n ,各类包含的实例数为m l (l =1,2;2l =1m l =m ),两类实例子集关于特征集合U n 一致,即X =X 1∪X 2,且X 1∩X 2=Ø.采用GA 方法实现P2P 和非P2P 流量两类实例的特征子集选择,染色体采用二进制编码:s :a 1a 2...a n ,其中a j 对应的是流统计特征属性,如分组到达时间间隔,a j =1或0表示第j (j =1,2,...,n )个属性包含或不包含在新的特征子集中.适应值函数设定为f (s )= m ln 2+m k =12 l =1P (X l |e k ,s )×ln(P (X l |e k ,s ))+α×(n −n j =1x j )(1)式中,α为熵函数值与特征子集简单性之间的平衡系统,在保证较小的错误率(反映P2P 流量的识别精度)的情况下,一般为1 α m/n [15].1.2分类算法贝叶斯分类是基于随机事件概率分布的统计分类方法,其中朴素贝叶斯分类是一种有监督的学习方法,它假定一个属性的值对给定类的影响独立于其他属性的值.而贝叶斯网络是一种图形模型,能够表示属性子集之间的依赖.1.2.1朴素贝叶斯分类朴素贝叶斯分类模型将事件的先验概率与后验概率巧妙地联系起来,利用先验信息和样本数据信息确定事件的后验概率,是目前公认的一种简单而有效的概率分类方法,已被应用到网络流量分类之中[10].用X 表示样本,S 表示所有样本的集合.A ={A 1,A 2,...,A n }表示样本集合S 上定义的n 个特征值,样本可以写成X ={x 1,x 2,...,x n },其中x 1,x 2,...,x n 分别是对应特征A 1,A 2,...,A n 的值;类变量C 的取值范围为{c j |1 i m,是正整数},在本命题中是两类:P2P 与非P2P(即m =2).给定样本X 属于类别c j 的极大后验概率可表示为c MAP =arg max j(p (c j |X ),1 j m(2)即表示在观测到的n 维特征A 条件下,类别c j 的所有条件概率中最大者为其归属的类.由贝叶斯公式,有c MAP =arg max j(p (c j |X ))=arg maxjp (X |c j )p (c j )p (X )=arg max jp (X |c j )p (c j )(3)式中,p (c j )可以从训练数据中由频率估算得到,此时需要计算p (X |c j ).朴素贝叶斯分类是基于一个简单的假定:所有的属性变量之间条件相互独立.则联合概率p (x 1,x 2,...,x n |c j )为p (x 1,x 2,...,x n |c j )=ip (x i |c j )(4)则c MAP =arg max jp (c j )ip (x i |c j )(5)相对于其他的贝叶斯分类方法,朴素贝叶斯分类算法的最大特点是不需要搜索,只需简单地计算训练实例中各个属性值发生的频数,就可以估计出每个属性的概率估计值.选择类c j 的所有条件概率中的最大者为其归属的类,该方法可获得很高的效率.但是,朴素贝叶斯分类要求A ={A 1,A 2,...,A n }的各个属性之间完全独立,这在网络流量分类中是不可能完全满足的,因此影响了分类精度.1.2.2贝叶斯网络朴素贝叶斯分类要求所有属性变量条件独立假定过于严格,在很多情形下都无法成立.贝叶斯网络则提供了一种折中的方法,它是假定在一个子集上的条件独立,这比朴素贝叶斯分类全局条件独立假定的限制要少,又比在所有变量中计算条件依赖更可行,适用于P2P 流量识别.贝叶斯网络是用一组条件概率函数以有向无循环图(DAG )形式表示不确定性的因果推理模型.贝叶斯网为表示联合概率分布,指定一组条件独立性假设(表示为DAG)以及一组局部条件概率集合.网络属性变量的元组{A 1,A 2,...,A n }分别取值为{x 1,x 2,...,x n }的联合概率为P (x 1,x 2,...,x n )=P (x n |x n −1,...,x 1)·P (x n −1|x n −2,...,x 1)...P (x 2|x 1)P (x 1)=P (x 1)n i =2P (x i |x i −1,...,x 1)(6)在贝叶斯网中,每一节点在给定其父母节点后都条件独立于它的前辈节点.则P (x 1,x 2,...,x n )=P (x 1)n i =2P (x i |M parents (A i ))(7)式中,M parents (A i )表示网络中A i 的父母节点的集合,ni =2P (x i |M parents (A i ))的值为与节点A i 关联的条件概率表中的值.第2期李君等:基于贝叶斯网络的Peer-to-Peer识别方法1271.2.3贝叶斯网络的学习评估函数的定义和搜索空间中搜索方法的确定,为贝叶斯网络建立了一种学习算法.这种学习可视为最优化问题,采用启发搜索方法构造模型并用得分方法来评价,目的就是发现评分最大的结构.K2是一种简单而快速的学习算法,它起始于某个给定的属性排序,然后对每个节点依次进行处理,贪心地增加从先前处理过的节点指向当前节点的连线,其每一步都增加能使网络评分达到最高值的连线.另一种贝叶斯网络分类器是由Friedman和Goldszmidt提出的树增强型朴素贝叶斯网络(tree aug-mented Naive Bayes,TAN).TAN是一种树状结构的贝叶斯模型,在TAN结构中,类变量是根,没有父结点,即S C=Ø(S C表示C的父节点集),类变量是每个属性变量的父节点,即C∈S Ai(S Ai表示A i的父节点集,i=1,2,...,n).属性变量A i除了以类变量C作为其父节点外,最多有1个其他的属性变量作为其父节点.则式(7)可简化为P(x1,x2,...,x n)=P(x1)ni=2P(x i|x p,c j),x p∈{x1,x2,...,x i−1}(8)TAN放松了朴素贝叶斯中的独立性假设条件,扩展了朴素贝叶斯的结构,允许每个属性结点最多可以依赖于1个非类节点.TAN具有较好的综合性能,体现了学习效率与分类精度之间的适当折中.BAN(Bayesian network augmented Naive Bayes)进一步扩展了TAN的结构,它允许属性之间可以形成任意的有向图,使其表示依赖关系的能力增强.BAN模型原则上对每个结点的父结点个数没有限定,只需按照事先选定的评价准则,在与属性节点A i相关联的节点A1,A2,...,A i−1和C中寻找A i的父节点,分类变量C是所有属性变量的父节点,每个节点A i可以找出多个父节点.TAN和BAN都是通过放松朴素贝叶斯的独立性假设来改进分类器性能.BAN比TAN的独立性假设更弱,往往都是在完全图中对整个弧空间进行搜索,选择最佳的弧集.由于任意弧反映了两个属性之间的依赖关系,因此也能反映流统计特征中所存在的依赖关系,如流的总字节数依赖于流的分组数与分组长度.现有的学习方法还采用标准的启发式技术,如贪婪爬山算法、遗传算法和模拟退火等.2实验方法及结果分析2.1实验数据集许多研究流量分类的实验数据是通过净荷分析法或者端口识别法给每个分组打上类别标签而获得的,训练和测试集的业务类标签不精确,会影响分类模型的准确性和分类性能的评估.通过组建一个测试局域网,让各主机在同一时段分别运行不同的应用,包括HTTP, SMTP,POP3,FTP,BIOS,GAME,DNS,Streaming, QQ,KaZaa,Gnutella,eMule,BitTorrent,Skype等,按不同时段(如假期或非假期,以及每天不同的时段)模拟相应用户的行为,采集Internet分组存入数据库,获得原始数据.由于预先限定每台主机运行的应用,可以很容易地根据IP地址快速并精确地标上业务类型的标签,得到精确可靠的训练和测试集.因为完全按正常使用各类业务,所以抓取的各应用分组可以代表真实的Internet环境中的业务流.为了克服上述网络流量不是全业务的缺点,采集实际的南京邮电大学校园网和浙江万里学院校园网出口流量作进一步的试验和验证.采用分类精度相对较高的净荷匹配法,为采集的训练与测试数据预先分类并打上标签,作为分类基准,用来评估各分类算法.由于存在分类错误,所以不可避免地存在评估偏差,但并不妨碍验证.对上述获得的数据集作如下处理:1)分流:根据五元组(源地址、目的地址,源端口、目的端口和传输层协议)来划分流.对于TCP流,由TCP的三次握手来识别流的开始,由TCP的FIN/RST分组作为流结束的标志.此外,若连接持续空闲了t s(如90s),则假定该流已经结束.同样,对于UDP流,若t s(取90s)时间内没有分组到达,则认为该UDP流结束.2)统计流的相关特征如流持续时间、分组到达时间间隔、分组净荷长度等,详细参见2.2.1节.3)从流统计数据库中随机选取一定数量的样本,形成训练和测试集.2.2基于贝叶斯网络的P2P流量分类性能实验与分析分类性能由分类精度和分类代价决定,分类代价由分类时间决定.分类精度的评估测度如下:分类精度A precision=N TPN TP+N FP召回率A recall=N TPN TP+N FN分类精度和召回率是对每个类进行评估的,而分类器的总体精度可表示为A overall=ni=1N iTPni=1(N iTP+N iFP)其中,对某一给定类,N TP和N FP分别代表正确分类的样本数量和误分到该类的样本数量,一个类的样本被错误地分到另一类的样本数,称N FN.2.2.1特征属性选择由经验获得的特征属性集合S={流持续时间,源端口,目的端口,流的总分组数,流的总字节数,分组到达128应用科学学报第27卷时间间隔中值,分组到达时间间隔平均值,分组到达时间间隔方差,分组到达时间间隔最大值,分组到达时间间隔最小值,分组净荷平均值,分组净荷中值,分组净荷方差,分组净荷最大值,分组净荷最小值,分组大小平均值,分组大小中值,分组大小方差,分组大小最大值,分组大小最小值},作为实验的属性全集.采用遗传算法获取最优子集:选取19946个实例流,其中各种P2P 流为10166(包括Gnutella,KaZaa,Skype,BitTrent,eMule 等),而非P2P 流为9780(包括HTTP ,FTP ,BIOS ,POP3,QQ ,DNS 等).采用遗传算法的特征子集选择后得到的最优特征子集为S ={流持续时间,流的总字节数,分组到达时间间隔中值,分组到达时间间隔平均值,分组到达时间间隔最大值}.2.2.2不同机器学习算法的比较和分析本节比较了贝叶斯网络各算法与文献中常用的朴素贝叶斯分类算法和反向传播神经网络方法的实验结果.从2.1节制备的数据集中随机选取19946个实例流,特征属性集为2.2.1节中根据遗传算法求得的最优特征子集S ,有限分类集合C ={P2P ,NP2P}分别表示为P2P 业务流和非P2P 业务流,采用十折交叉验证方法进行实验.从表1的实验结果可见,反向传播(BP)神经网络方法和朴素贝叶斯分类方法把许多非P2P 流错误地划分为P2P 流,导致非P2P 流的N TP 下降,而N FN 数值大大增加,造成分类精度下降,召回率极低.贝叶斯网络分类算法的分类精度都能达到95%以上,明显高于朴素贝叶斯分类.再比较各贝叶斯网络学习算法,用遗传算法进行搜索的贝叶斯网络算法分类复杂度大,所需的分类时间达到21min 之多,模拟退火搜索算法竟超过1h ,而分类精度并不比TAN 和BAN 高.其他各类贝叶斯网络分类算法的速度基本上能达到Gbps 的数量级.用高速处理器替代实验所用的个人计算机,有可能在边缘路由器上实现P2P 业务实时识别,其中K2的分类速度最快且具有较高的分类精度.综合考虑分类性能与代价,K2,TAN 和BAN 算法的分类精度相对较好且所需的分类时间较少,是比较理想的分类算法.表1各种分类算法的比较Table 1Comparison of the classifiers分类算法A precision A recall A overall分类时间/s参数说明NP2P P2P NP2P P2P 朴素贝叶斯0.2920.6940.0210.9780.6852核估计反向传播神经网络0.9860.6970.015 1.0000.6994961个隐层贝叶斯网络(K2)0.9300.9590.9070.9700.95171个父节点贝叶斯网络(爬山)0.9300.9590.9070.9700.951141个父节点贝叶斯网络(遗传算法)0.9480.9600.9080.9780.95712611个父节点贝叶斯网络(模拟退火)0.9520.9600.9080.9800.95836231个父节点贝叶斯网络(TAN)0.9470.9630.9150.9780.959142个父节点贝叶斯网络(BAN)0.9530.9650.9190.9800.96114,18,19(respectively)3,4,5个父节点2.2.3采用优选子集与属性全集的贝叶斯网络分类性能比较采用2.2.2节得到的分类性能较好的贝叶斯网络算法:K2,TAN,BAN ,分别比较在优选属性子集S 和属性全集S 上的分类精度和分类时间,如表2∼4所示.从以上实验可知,在属性全集上训练的分类方法增加了计算复杂度,分类时间成倍增长.同时由于属性全集中存在无用或冗余的特征属性,影响了学习质量,导致分类精度下降,在K2贝叶斯网络分类中尤为明显.基于属性全集的TAN 的总体精度也不如优选子集的总体精度,BAN 虽然在属性全集上的总体精度略高于优选子集,但分类时间超过优选子集所需分类时间的3倍.所以特征选择在保证分类精度基本保持不变甚至有所提高的情况下,能够大大降低分类复杂度,同时也证明了基于遗传算法的特征选择在本例中的有效性.表2TAN 分类法在不同属性集上的性能比较Table 2Performance comparison of TAN al-gorithm using different feature sets 特征属性集A precision A recall A overall分类时间/s NP2P P2P NP2PP2P全集S 0.9240.9690.9310.9660.95547优选属性子集S0.9470.9630.9150.9780.95914第2期李君等:基于贝叶斯网络的Peer-to-Peer识别方法129表3有3个父节点的BAN分类法在不同属性集上的性能比较Table3Performance comparison of BAN algorithm(with3father nodes)using different feature sets特征属性集A precision A recallA overall分类时间/s NP2P P2P NP2P P2P全集S0.9640.9670.9250.9850.96648优选属性子集S0.9530.9650.9190.9800.96114表4K2贝叶斯网络分类算法在不同属性集上的性能比较Table4Performance comparison of K2al-gorithm using different feature sets特征属性集A precision A recallA overall分类时间/s NP2P P2P NP2P P2P全集S0.8980.9650.9220.9540.94429优选属性子集S0.9300.9590.9070.9700.95172.2.4未知P2P流量的识别由于新的P2P业务层出不穷,现有P2P业务升级频繁,同时为规避被识别,P2P协议更新也很频繁,这就要求一个好的P2P业务分类器能够识别未知P2P业务,且具有较好的分类鲁棒性和可扩展性.本实验采集一种在原训练集中未曾出现过的P2P业务流,与非P2P业务流混合形成测试集,测试样本数分别为9450和2076.采用前述的分类模型——基于优选子集的TAN分类模型,在该测试集上测试,并评估分类器的精度,目的是验证分类器识别未知P2P业务的能力.从表5可知,建立的分类器可以精确地识别未知P2P业务.表5未知P2P业务分类结果Table5Classification results for unknown P2P traffic样本数A precision A recallA overall分类时间/s NP2P P2P NP2P P2P94500.9430.9850.9430.9850.976420760.8770.9990.9340.9430.93922.2.5P2P流量的早期检测与实时应用当流结束时识别出该流所关联的业务对于很多网络应用意义都不大,尤其是对于网络监测、QoS保障等实时应用.P2P流量占据了大量的带宽,很多企业甚至ISP采用对P2P业务限流、整形和封阻的方法以保障关键业务的服务质量,这就要求能够早期检测P2P流量并能实时应用.Bernaille et al.[16]提出了仅依赖流的前5个分组,采用K-Means聚类算法来完成流量分类,达到了令人满意的结果,克服了基于机器学习的流量识别滞后的问题.同样,本文对流的前4∼8个分组进行分析,采用贝叶斯网络学习法识别P2P流量,实验结果如图2所示.可见,各贝叶斯网络学习算法只采用流的少量分组的统计值,就足以达到较高的分类精度,大部分能达到90%以上.同时,这种早期检测的方法可以减少流状态的维护和计算复杂度,减轻存储压力,有利于实时的网络监测与管理.packet number NAprecision/%图2分类精度vs.用于分类的分组数Figure2Classification accuracy vs.appliedpacket number perflow3结语由于Internet网络业务的多样性和复杂性,以及不少新业务(尤其是P2P文件共享业务)为了防止被识别和流量限制,采用了随机动态端口和伪装端口,甚至改变一些可能被识别的流量特性,给网络业务的智能识别与分类带来了很大的困难.通过相关的网络流量测试和分析,发掘流量特征属性,利用遗传算法选取最优特征子集,并采用TAN和BAN等贝叶斯网络分类算法进行P2P流量识别.实验结果表明,相对于朴素贝叶斯方法和BP神经网络方法,本识别系统在识别精度和识别速度上都有较大的优势,其中总体识别精度能达到95%以上.本识别方法还具有可扩展性,可用来识别未知的P2P流量,并能实现P2P业务的早期检测和实时识别.参考文献:[1]S en S,J ia W.Analyzing peer-to-peer traffic across largenetworks[J].IEEE/ACM Transactions on Networking,2004,12(2):219-232.[2]S aroiu S,G ummadi P K,D unn R J,G ribble S D,L evy H M.An analysis of internet content delivery sys-tems[C]//5th Symposium on Operating Systems Designand Implementation,2002.[3]K aragiannis T,B roido A,B rownless N,C laffy K,F aloutsos M.File-sharing in the internet:a character-ization of P2P traffic in the backbone[R].University ofCalifornia,Riverside Department of Computer Science,Nov.2003.[4]S en S,S patscheck O,W ang D.Accurate,scalablein-network identification of P2P traffic using applica-tion signatures[C]//Proceedings of the13th Interna-130应用科学学报第27卷tional Conference on World Wide Web,New York,2004: 512-521.[5]K aragiannis T,B rodio A,F aloutsos M,C laffy K.Transport layer identification of P2P traffic[C]//Pro-ceedings of the4th ACM SIGCOMM Conference on In-ternet Measurement,Taormina,2004:121-134.[6]K aragiannis T,P apagiannaki K,F aloutsos M.Blinc:multilevel traffic classification in the dark[C]// Philadelphia,Pennsylvania:SIGCOMM’05,2005. [7]C onstantinou F,M avrommantis P.Identifyingknown and unknown peer-to-peer traffic[C]//Proceed-ings of Fifth IEEE International Symposium on Net-work Computing and Applications(NCA2006),Cam-bridge,Massachusetts,July24-26,2006:93-102.[8]Z ander S,N guyen T,A rmitage G.Automated traf-fic classification and application identification using ma-chine learning[C]//Proceedings of the IEEE30th Con-ference on Local Computer Networks(LCN2005),Syd-ney,November2005:250-257.[9]E rman J,A rlitt M,M ahanti A.Traffic classifica-tion using clustering algorithms[C]//Proceedings of SIGCOMM Workshop on Mining Network Data,Pisa, September11-15,2006:281-286.[10]M oore A W,Z uev D.Internet traffic classification us-ing Bayesian analysis techniques[C]//Proceedings of the2005ACM SIGMETRICS International Conference on Measurement and Modeling of Computer Systems, Banff,Alberta,June6-10,2005:50-60.[11]J iang H,M oore A W,G e Z.Lightweight applica-tion classification for network management[C]//Pro-ceedings of the2007SIGCOMM Workshop on Internet Network Management,Kyoto,2007:299-304.[12]W ang R,L iu Y,Y ang Y.Solving the app-level clas-sification problem of P2P traffic via optimized support vector machines[C]//Proceedings of Sixth International Conference on Intelligent Systems Design and Applica-tions,Jinan,2006:534-539.[13]A uld T,M oore A W,G ull S F.Bayesian neural net-works for internet traffic classification[J].IEEE Trans-actions on Neural Networks,2006,18(1):223-239. [14]M oore A W,Z uev D.Discriminators for use inflow-based classification[R].Intel Research,Cambridge, 2005.[15]李敏强,寇纪淞,林丹.遗传算法的基本理论与应用[M].北京:科学出版社,2002:297-310.L i Minqiang,K ou Jisong,L in Dan.Theory and applica-tion of genetic algorithm[M].[S.l.]:Science Press,2002: 297-310.(in Chinese)[16]B ernaille L,T euxeira R,A kodkenous I,S oule A,S lamatian K.Traffic classification on thefly[J].ACM SIGCOMM Computer Communication Review,2006, 36(6):23-26.(编辑:欧阳丽霞)。

相关文档
最新文档