ArticlePDF Available

Intelligent Eco Networking (IEN): Knowledge-Driven and Value-Oriented Future Internet Infrastructure (智能生态网络:知识驱动的未来价值互联网基础设施)

Authors:

Abstract and Figures

As the trend of knowledgelization of content, evaluation of knowledge, networking of value, ecologicalization of network and intellectualization of ecology is becoming increasingly prominent in the future Internet. In this paper, we propose the concept of intelligent eco networking (IEN) for the future Internet, countering with the deficiencies of the current IP networks, including rigid architecture, weak content awareness, poor multi-architecture/multi-network integration, low scheduling flexibility, lacking endogenous security and trust maintenance mechanism, single quality of service (QoS) mode and outdated evaluation indicators and methods. IEN adopts software and hardware integrated technology roadmap based on virtualization and configurable devices, improves information-centric networking (ICN) architecture, integrates distributed artificial intelligence (AI) analysis decision and blockchain consensus computing technology, considers network resource cost/profit indicators regarding storage, computing and bandwidth resources, aims at building a hierarchical, intelligent and semantic network architecture. IEN is backward compatible with IP protocol; forward evolves towards naming and IP integrated, heterogeneous computing addressing and multi-modal network protocol for cross-domain, edge-critical scenario, overlays content, identity authentication and multi-party trusted incentive mechanism; optimizes network resources allocation model and evaluation system. Through content semantic detection and identity credibility authentication, IEN insists on both security and openness. IEN aims to form a network infrastructure with high expansion, dynamic adaptation, and multi-objective optimization. It explores a new generation of industrialization, economics, and ecological Internet, and establishes an intelligent network of openness, sharing and synergy.
Content may be subject to copyright.
38 1
20201
应 用 学 学
JOURNAL OF APPLIED SCIENCES Electronics and Information Engineering
Vol.38 No. 1
Jan. 2020
DOI: 10.3969/j.issn.0255-8297.2020.01.012
动的设施
1,2
1
1
1
英英1
2
1. 学 信518055
2. 518055
.IP/
调度护机
新性intelligent eco networking, IEN. IEN
线
存储/
新型. IEN IP
命名IP 的多
.
IEN可控个高
设施砥砺奠定
互惠.
命名
TP399 0255-8297(2020)01-0152-21
Intelligent Eco Networking (IEN): Knowledge-Driven and
Value-Oriented Future Internet Infrastructure
LEI Kai1,2, HUANG Shuokang1, FANG Junjie1, HUANG Jiyue1,
XIE Yingying1, PENG Bo2
1. Shenzhen Key Lab for Information Centric Networking &Blockchain Technology,
School of Electronics and Computer Engineering, Peking University,
Shenzhen 518055, China
2. Internet Research Institute, Peking University, Shenzhen 518055, China
Abstract: As the trend of knowledgelization of content, evaluation of knowledge, net-
working of value, ecologicalization of network and intellectualization of ecology is becom-
ing increasingly prominent in the future Internet. In this paper, we propose the con-
稿2019-10-31
改高 [2016]2533
No.ZDSYS201802051831427
命名. E-mail: leik@pkusz.edu.cn
1动的设施 153
cept of intelligent eco networking (IEN) for the future Internet, countering with the defi-
ciencies of the current IP networks, including rigid architecture, weak content awareness,
poor multi-architecture/multi-network integration, low scheduling flexibility, lacking en-
dogenous security and trust maintenance mechanism, single quality of service (QoS) mode
and outdated evaluation indicators and methods. IEN adopts software and hardware in-
tegrated technology roadmap based on virtualization and configurable devices, improves
information-centric networking (ICN) architecture, integrates distributed artificial intelli-
gence (AI) analysis decision and blockchain consensus computing technology, considers net-
work resource cost/profit indicators regarding storage, computing and bandwidth resources,
aims at building a hierarchical, intelligent and semantic network architecture. IEN is back-
ward compatible with IP protocol; forward evolves towards naming and IP integrated,
heterogeneous computing addressing and multi-modal network protocol for cross-domain,
edge-critical scenario, overlays content, identity authentication and multi-party trusted in-
centive mechanism; optimizes network resources allocation model and evaluation system.
Through content semantic detection and identity credibility authentication, IEN insists on
both security and openness. IEN aims to form a network infrastructure with high expan-
sion, dynamic adaptation, and multi-objective optimization. It explores a new generation
of industrialization, economics, and ecological Internet, and establishes an intelligent net-
work of openness, sharing and synergy.
Keywords: future Internet, named data networking (NDN), knowledge-driven, blockchain,
federated learning (FL), heterogeneous computing network
万物沿.
仅局
新兴.
的多同体
.3存储
[1].
.
[2]
Internet of Things, IoT[3]
edge computing
.5G6G
IP 12/
3.
IP
.新兴IP
/
.IP
护机在云.
IP quality of service, QoSBest Effort Only
.
IP 调度[4]
artificial intelligence, AI
.应用术使
.势是解决[5].
[6-7][8] [9] 高复能能.
154 应 用 学 学 38
调度[10][11]实时
.
能模
更高.
.早在 20 90 由于助智
distributed artifical intelligence, DAI
[12]. DAI 解决
解决.各个
.应用
型训
时受于用.
[13] 新型
——federal learning, FL.1
据进行训点的
2各个直至主.
得到定的.
对当兴需
新型——intelligent eco networking, IEN
[14].
1.
解决IP .
2为网
/.
3.
4.
QoS 解决
.
1
TCP/IP
国国5
命名(named data networking, NDN)
新型.
1.1
TCP/IP .
IoT [15]
.
道的分发
.构和
Best Effort Only
1动的设施 155
应用.
TCP/IP .
1定导架僵.IP
黑盒.SDN
用于
.
2.件技IP
黑盒现象.致重
就近的动
端到端的应用
.
.
3. IP IP
.
.得到IP
location based service, LBS [16].
5G的多步并.
4度低. IP 并不
.IP
IP IPSec VPN 解决
. IP 断的
道的.
.入认方法用于.
方法.段对
.2017 5
[17]IP
露了方法
IP 追踪.
1.2
TCP/IP
.NDN[18][19]information centric network,
ICN. NDN
分发.多的NDN 应用于新兴
NDN 用于IoT[20][3] . NDN IP为未
.
5
NDN[18] MobilityFirst [21] NEBULA [22] eXpressive Internet architecture [23](XIA)
ChoiceNet [24]. NDN 命名
. MobilityFirst
1探讨. NEBULA 聚焦
156 应 用 学 学 38
存储. XIA
可靠. ChoiceNet
.
NDN 种主
IP 为网分而使数
解决IP NDN
NDN 求强
NDN 实施由与发分
.
NDN 10NDN 仅仅
应用.NDN 仍然
NDN .
NDN 的底
动的
.NDN5.
1容缺.由于 NDN 分发
NDN
[25]NDN
.
2应用.点的
义以.
[26]NDN
.
3.
.
.
法反
.
4访护机.
.
NDN .
5泛分.
使NDN 仅仅
.也一低等.
[27]
BlockNDN 立类广
现象.
1.3
键技
设施应用.
1动的设施 157
黑盒署时
TCP/IP
应用
应用
远远超出应用
.些新兴
入人为网.
融入个更
.12
盾的3
4宏观解决
412
输数34线.
2
intelligent eco networking, IEN
动的设施. IEN
线存储Token
互惠1.
1
Figure 1 Network intelligence ecologicalization
2.1 heterogeneous computing network
.
备不.
.5G
158 应 用 学 学 38
存储.存储存
.得到.
究节点的
.点的
据结
识是DPDK
访RDMAP4 .
IP
.拥有++使
.
2.2 valuable content centric
户获意义
效信.应用成长
应用广.
意义
设施
意义.于用并不
.分发存储
的对.
万物广
.
1.
多的
.之中
接进便.使
等多.
2.IP .
道的不便.
实时线便.
.
使
.存储.
.可靠
使数拥有.
3确认权.
拥有.IP
确权..
使.于用
权认乏分
.
1动的设施 159
4便.
买卖.
于又.
设施.
点的.
IEN .
新型的底设施IEN
使.据均
确权.
地定. IEN
买卖
.
.
态体互惠.
IEN 设施
请求分发.定的
.IEN 聚焦机或
请求信效
. IEN 别并
设施供更.
2.3 knowledge-driven intelligence
IEN 使动的
. IP
耗和
.IEN
线.互惠IEN 聚焦
而非单纯.
3
用于.
由于.
和共是数字转.快控
证整的调度接近.应用
制智
改革.
2.4 authorization and authentication
IEN 使
.IP
160 应 用 学 学 38
端到端道的.
的多IP 应用
.
而非定的解决
确权.
设施
.
.
[28] .
2.5 (multi-party consensus)
IEN .
.
以依确权将价
.
设施..度的
解决确权.
淘汰
..
/
互惠.
实是.
点的
实施.
P2P .
P2P 解决
.
1使上去P2P 络逻
2点的
3IP P2P 实数
4实施可靠.
使IEN
.
1动的设施 161
2 IEN
Figure 2 Design and composition of IEN
3
3.1 (organizational structure)
3.1.1
3IEN .
AI 据结地调
解决.
该更便. IEN AI 点的使
.
3
Figure 3 Hierarchical heterogeneous computing network structure
162 应 用 学 学 38
IEN .
方法
.存储力来
接近端的.
3.1.2 会化
在这.
方访访[29].万物
仅仅局的对
..在这
由于仅仅万物[31].在这
繁而
.
广使TCP/IP
.TCP/IP
.直整——
[32]
.
TCP/IP 和管.
新兴——sofeware defined network, SDN
换机
[33].SDN
.
多的新兴
[34]
让全
.
3.1.3 动调
IEN 应用
制智
.4IEN
.
IEN 动调设施——
.便
析网.1
NDN 包包
匹配度等使
的调.的第 2是实
.方法
.
1动的设施 163
4动调
Figure 4 Data-driven regulation strategy running logic
3.2 (compatibility and evolution)
IP 现向IEN 术设.
3.2.1
IEN IP 使IP
1命名
命名使
IP 得到解决.
2
IP 这在
.
.
3
线.
TCP/IP .
使是什whatwhere
解决盾的方法[35] .
方法
1
TCP/IP NDN
.公告请求
广广便
请求.
2
164 应 用 学 学 38
TCP/IP 式是点的定的. NDN 式是
定的.使和哈命名
消息命名使TCP/IP NDN
.
3与硬件加
线IEN 现向.
. IEN
线
实世由硬
存储使
频谱
.此单
个更解决.
行协线IEN NDN
IP IP 方法调度
.NDN IEN
NDN IEN IP 自组
IEN
.
3.2.2
从传IEN IEN //.
存储.
面面.多的了满足自
存储. IEN //[36] 级计存储
为物网网存储.[37]
存储IEN IEN .
供更.接近
IEN .//
便供更低端的.//
.
.
3.3 (indicators and systems)
3.3.1 新型
IEN 动的
用优新型
.有用
.IEN
新型
1动的设施 165
容冗度的.
使3新型
可靠.IP NDN
ICN IEN 5.
5 IEN
Figure 5 Expected increased functional strength of IEN
3.3.2
和共段的.的到
多的.
.产产有与
存储.仅仅
.
.
.
.
IEN 为网点的
.
实时
.
.
确权将价
..度的解决确权
.
4行性及技线
4度对行性
行性
IEN .IEN 种逐设施.
166 应 用 学 学 38
4.1 from the edge to the core
.动的
5G
新型
.
[39].新型
践价.
访吞吐
.络流使
[40].
6示深络流
IPv4IPv6线WiFi
7城出
400 M5 h 400 M
500 M.
62018 6
Figure 6 Network topology of Shenzhen University Town (June 2018)
IEN .
1络流调度
SDN 调度
络流调度.
1动的设施 167
72018 11 路流实时
Figure 7 Real-time dynamics of the export link traffic of the education network in Shenzhen
University Town in November, 2018
2调度
调度的CDN NDN
习相调度
.
3调度
NDN 各个点独致整
.NDN
.
4
调度
jitter 抖动
.
4.2 value-oriented consensus
定的
据交
.往往是属意义
. IEN 式身[41]distributed identification, DID
8.
IEN 建价.
过共. IEN
确权使据交使
度地设施
存储.
168 应 用 学 学 38
8式身
Figure 8 Blockchain-based distributed identification
4.3 distributed ID authentication
IEN .NDN TCP/IP
新型.在这使
匹配NDN 沿反方.
.动的术使
[18].
NDN 广使为网
.
实数.NDN
由于.
应用及检
.于有坏基
NDN 可靠.
解决51
23
毒的45
实施.使
链来钥摘publisher public key digest, PPKD
--interest-key-content binding, IKCB.
PPKD
钥摘
.使毒的仍然存处
的多
.
使[42] .
4.4
新兴式深使
1动的设施 169
者在拥有
式数.IEN
.应用[43]
都得
据进行训度的. 1
据具定的
2得到的3
是收
general data protection regulation, GDPR在这
[44].在这
解决
得到[45]
.
行训
广点对
[46].
.
端都.使
何贡.
.
些研低低端对.
点的用于
型性
像现在这.
点的
想相
点的
.端的使
更高.
低低解节
均具要意义.
.合和也引.
定的
仅仅或汇
端的.得到实施
络流.
单纯
据进.
170 应 用 学 学 38
4.5
IEN
.IEN
1.ICN IEN 正逐
存储奠定.
据进IEN 动的.
2动的. IEN 动的
.
的调IEN .
3. IEN
IEN .
4.IEN
QoS 与语访靠可
QoS .
5
IEN
.互惠
仅仅分发
. IEN
互惠.
IEN 多的.
IEN 为未在这QoS
调的.
学性
. IEN 为未设施
.
:
[1] Peterson L L, Davie B S. Computer networks: a systems approach [M]. [S.l.]: Elsevier, 2007.
[2] Atzori L, Iera A, Morabito G. The Internet of Things: a survey [J]. Computer Networks,
2010, 54(15): 2787-2805.
[3] Shi W, Cao J, Zhang Q, et al. Edge computing: vision and challenges [J]. IEEE Internet of
Things Journal, 2016, 3(5): 637-646.
[4] Nilsson N J. Principles of artificial intelligence [M]. San Francisco: Morgan Kaufmann, 2014.
[5] Russell S J, Norvig P. Artificial intelligence: a modern approach [M]. Malaysia: Pearson
Education Limited, 2016.
1动的设施 171
[6] Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional
neural networks [C]//Advances in Neural Information Processing Systems, 2012: 1097-1105.
[7] He K, Zhang X, Ren S, et al. Deep residual learning for image recognition [C]//Proceedings
of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770-778.
[8] Hinton G, Deng L, Yu D, et al. Deep neural networks for acoustic modeling in speech
recognition [J]. IEEE Signal Processing Magazine, 2012: 29.
[9] Bahdanau D, Cho K, Bengio Y. Neural machine translation by jointly learning to align and
translate [OL]. 2014. arXiv Preprint arXiv:1409.0473.
[10] Kim H, Gupta A. Ontas: flexible and scalable online network traffic anonymization system
[C]//Proceedings of the 2019 Workshop on Network Meets AI & ML. ACM, 2019: 15-21.
[11] Su K, Li J, Fu H. Smart city and the applications [C]//2011 International Conference on
Electronics, Communications and Control (ICECC). IEEE, 2011: 1028–1031.
[12] Chaib-Draa B. Readings in agents [M]. San Francisco: Morgan Kaufmann, 1998.
[13] Smith V, Chiang C K, Sanjabi M, et al. Federated multi-task learning [C]//Advances in
Neural Information Processing Systems, 2017: 4424-4434.
[14] Turing S. Intelligent eco networking (IEN): an advanced future Internet of intelligence for digi-
tal social economic ecosystem [C]//2018 1st IEEE International Conference on Hot Information-
Centric Networking (HotICN). IEEE, 2018: 179-185.
[15] Shang W, Yu Y D, Droms R, et al. Challenges in IoT networking via tcp/ip architecture
[R]. Technical Report NDN-0038. NDN Project, 2016.
[16] Rankin P J, Griffiths J C. Distributed location based service system: US 6879838 [P].
2005-12.
[17] Mohurle S, Patil M. A brief study of Wannacry threat: Ransomware attack [J]. International
Journal of Advanced Research in Computer Science, 2017, 8(5).
[18] Zhang L, Afanasyev A, Burke J, et al. Named data networking [J]. ACM SIGCOMM
Computer Communication Review, 2014, 44(3): 66-73.
[19] Ahlgren B, Dannewitz C, Imbrenda C. et al. A survey of information-centric networking
[J]. IEEE Communications Magazine, 2012, 50(7): 26-36.
[20] Shang W, Bannis A, Liang T, et al. Named data networking of things [C]//2016 IEEE
First International Conference on Internet-of-Things Design and Implementation (IoTDI). IEEE,
2016: 117-128.
[21] Raychaudhuri D, Nagaraja K, Venkataramani A. Mobilityfirst: a robust and trustworthy
mobility-centric architecture for the future internet [J]. ACM SIGMOBILE Mobile Computing
and Communications Review, 2012, 16(3): 2-13.
[22] Anderson T, Birman K, Broberg R, et al. The nebula future internet architecture [M]//The
Future Internet Assembly. [S.l.]: Springer, 2013: 16-26.
[23] Anand A, Dogar F, Han D, et al. Xia: an architecture for an evolvable and trustworthy
Internet [C]//Proceedings of 10th ACM Workshop on Hot Topics in Networks. ACM, 2011: 2.
[24] Wolf T, Griffioen J, Calvert K L, et al. Choicenet: toward an economy plane for the
internet [J]. ACM SIGCOMM Computer Communication Review, 2014, 44(3): 58-65,
[25] Yuan H, Song T, Crowley P. Scalable NDN forwarding: concepts, issues and principles
[C]//2012 21st International Conference on Computer Communications and Networks (ICCCN).
IEEE, 2012: 1-9.
[26] Lei K, Qin M, Bai B, et al. GCN-GAN: a non-linear temporal link prediction model for
weighted dynamic networks [C]//IEEE INFOCOM 2019-IEEE Conference on Computer Com-
munications. IEEE, 2019: 388-396.
[27] Jin T, Zhang X, Liu Y, et al. Block NDN: a bitcoin blockchain decentralized system over
named data networking [C]//2017 Ninth International Conference on Ubiquitous and Future
Networks (ICUFN). IEEE, 2017: 75-80.
[28] Zhou Z, Li L, Li Z. Efficient cross-domain authentication scheme based on blockchain tech-
nology [J]. Journal of Computer Applications, 2018, 38(2): 316-320.
172 应 用 学 学 38
[29] Nunes B A A, Mendonca M, Nguyen X N, et al. A survey of software-defined network-
ing: past, present, and future of programmable networks [J]. IEEE Communications Surveys &
Tutorials, 2014, 16(3): 1617-1634.
[30] .[J]. 线2017, 47(11): 17-21.
Han Q M. Research on architecture of intelligent service system about man-machine-object
interactive network [J]. Radio Engineering, 2017, 47(11): 17-21. (in Chinese)
[31] Kreutz D, Ramos F M V, Verissimo P E. et al. Software-defined networking: a compre-
hensive survey [J]. Proceedings of the IEEE, 2014, 103(1): 10-13.
[32] Rothenberg C E, Nascimento M R, Salvador M R, et al. Revisiting routing control
platforms with the eyes and muscles of software- defined networking [C]//ACM SIGCOMM
Workshop on Hot Topics in Software Defined Networking, 2012.
[33] Li X, Jiang P, Chen T, et al. A survey on the security of blockchain systems [J]. Future
Generation Computer Systems, 2017. DOI:10.1016/j.future.2017.08.020
[34] Jacobson V, Smetters D K, Thornton J D, et al. Networking named content
[C]//Proceedings of the 5th International Conference on Emerging Networking Experiments
and Technologies. ACM, 2009: 1-12.
[35] Josep A D, Katz R, Konwinski A, et al. A view of cloud computing [J]. Communications of
the ACM, 2010, 53(4).
[36] Bonomi F, Milito R, Zhu J, et al. Fog computing and its role in the Internet of Things
[C]//Proceedings of the First Edition of the MCC Workshop on Mobile Cloud Computing.
ACM, 2012: 13-16.
[37] .SDN 应用 [J]. 计技2014(3) :
35-39.
Guo A P, Zhou G T, Xia J J, et al. Edge network control technology based on SDN and
its application [J]. Designing Techniques of Posts and Telecommunications2014(3) : 35-39. (in
Chinese)
[38] 郑植.校校[J]. 息系2013, 12: 42.
[39] Mattila J. The blockchain phenomenon-the disruptive potential of distributed consensus ar-
chitectures [R]. ETLA Working Papers, 2016.
[40] Lei K, Zhang Q, Lou J, et al. Securing ICN-based UAV ad hoc networks with blockchain [J].
IEEE Communications Magazine, 2019, 57(6): 26-32.
[41] Wu X, Kumar V, Quinlan J R, et al. Top 10 algorithms in data mining [J]. Knowledge and
Information Systems, 2008, 14(1): 1-37.
[42] Mcmahan H B, Ramage D, Talwar K, et al. Learning differentially private recurrent lan-
guage models [J/OL]. 2017. arXiv preprint arXiv:1710.06963
[43] Yang Q, Liu Y, Chen T, Tong Y. Federated machine learning: concept and applications [J].
ACM Transactions on Intelligent Systems and Technology, 2019, 10(2): 1-19.
[44] Xie P, Bilenko M, Finley T, et al. Crypto-nets: neural networks over encrypted data [J/OL].
2014.arXiv preprint arXiv:1412.6181.
()
ResearchGate has not been able to resolve any citations for this publication.
Conference Paper
Full-text available
Access to packet traces is required not only to detect and diagnose various network issues related to performance and security, but also to train intelligent learning models enabling networks that can run themselves. However, packets in a network carry a lot of information which can be used to personally identify users and their online behavior. This requires network operators to anonymize packet traces before sharing them with other researchers and analysts. Existing tools anonymize packet traces in an offline manner, which incurs significant computational, storage, and memory overhead---limiting their ability to scale as the volume of the collected packet trace increases. In this paper, we present the design and implementation of an Online Network Traffic Anonymization System, ONTAS, which can flexibly anonymize packet traces in the data plane itself using modern PISA-based programmable switches.
Article
Full-text available
Since its inception, the blockchain technology has shown promising application prospects. From the initial cryptocurrency to the current smart contract, blockchain has been applied to many fields. Although there are some studies on the security and privacy issues of blockchain, there lacks a systematic examination on the security of blockchain systems. In this paper, we conduct a systematic study on the security threats to blockchain and survey the corresponding real attacks by examining popular blockchain systems. We also review the security enhancement solutions for blockchain, which could be used in the development of various blockchain systems, and suggest some future directions to stir research efforts into this area.
Conference Paper
Full-text available
Blockchain provides a new approach for participants to maintain reliable databases in untrusted networks without centralized authorities. It resolves the present problem that all the deals and cooperations between strangers have to rely on trusted third parties. Meanwhile, blockchain makes it possible that a network of completely homogeneous nodes can reach consensus. However, there are still many serious problems in real blockchain systems in IP network such as the lack of support for multicast and the possibility of the existence of supernodes. In this paper, we design and implement a bitcoin-like blockchain system over Named Data Networking (NDN), a promising future network architecture, to find out if it can solve the problems mentioned above. By using our naming rules, nodes fetch data from individual participants and send out self-state information at the same time, completing the synchronization of massive dataset. Nodes can search the specified data with rules as well. The resulting design solves those problems in IP network. It provides completely decentralized systems and simplifies system architecture. It also improves the weak-connectivity phenomenon and decreases the broadcast overhead. It’s worth pointing out that our ingenious system can work as an infrastructure of NDN architecture to provide services in different application systems. Keywords—blockchain; NDN; peer-to-peer; bitcoin; next generation networks
Article
Recently Ransomware virus software spread like a cyclone winds. A cyclone wind creates atmospheric instability; likewise ransomware creates computer data instability. Every user is moving towards digitization. User keep data secure in his or her computer. But what if data is hijacked. A ransomware is one of the software virus that hijack users data. A ransomware may lock the system in a way which is not for a knowledgeable person to reverse.It not only targets home computersbut business also gets affected. It encrypts data in such a way that normal person can no longer decrypt. A person has to pay ransom to decrypt it. But it does not generate that files will be released. This paper gives a brief study of WannaCry ransomware, its effect on computer world and its preventive measures to control ransomware on computer system.
Article
Today’s artificial intelligence still faces two major challenges. One is that, in most industries, data exists in the form of isolated islands. The other is the strengthening of data privacy and security. We propose a possible solution to these challenges: secure federated learning. Beyond the federated-learning framework first proposed by Google in 2016, we introduce a comprehensive secure federated-learning framework, which includes horizontal federated learning, vertical federated learning, and federated transfer learning. We provide definitions, architectures, and applications for the federated-learning framework, and provide a comprehensive survey of existing works on this subject. In addition, we propose building data networks among organizations based on federated mechanisms as an effective solution to allowing knowledge to be shared without compromising user privacy.
Article
Federated learning poses new statistical and systems challenges in training machine learning models over distributed networks of devices. In this work, we show that multi-task learning is naturally suited to handle the statistical challenges of this setting, and propose a novel systems-aware optimization method, MOCHA, that is robust to practical systems issues. Our method and theory for the first time consider issues of high communication cost, stragglers, and fault tolerance for distributed multi-task learning. The resulting method achieves significant speedups compared to alternatives in the federated setting, as we demonstrate through simulations on real-world federated datasets.