1 1 Detecting Network Neutrality Violations with Causal Inference Mukarram Bin Tariq, Murtaza Motiwala Nick Feamster, Mostafa Ammar {mtariq, murtaza, feamster,

Slides:



Advertisements
Similar presentations
TWO STEP EQUATIONS 1. SOLVE FOR X 2. DO THE ADDITION STEP FIRST
Advertisements

You have been given a mission and a code. Use the code to complete the mission and you will save the world from obliteration…
Advanced Piloting Cruise Plot.
Chapter 1 The Study of Body Function Image PowerPoint
Effective Change Detection Using Sampling Junghoo John Cho Alexandros Ntoulas UCLA.
Network Monitoring System In CSTNET Long Chun China Science & Technology Network.
A New Method for Symmetric NAT Traversal in UDP and TCP
Wenke Lee and Nick Feamster Georgia Tech Botnet and Spam Detection in High-Speed Networks.
NANO: Network Access Neutrality Observatory Mukarram Bin Tariq, Murtaza Motiwala, Nick Feamster, Mostafa Ammar Georgia Tech.
Wenke Lee and Nick Feamster Georgia Tech Botnet and Spam Detection in High-Speed Networks.
Detecting Network Neutrality Violations with Causal Inference Mukarram Bin Tariq, Murtaza Motiwala Nick Feamster, Mostafa Ammar Georgia Tech
1 Building a Fast, Virtualized Data Plane with Programmable Hardware Bilal Anwer Nick Feamster.
Nick Feamster Georgia Tech Joint work with Mukarram bin Tariq, Murtaza Motiwala, Yiyi Huang, Mostafa Ammar, Anukool Lakhina, Jim Xu Detecting and Diagnosing.
Access Networks: Applications and Policy Nick Feamster CS 6250 Fall 2011 (HomeOS slides from Ratul Mahajan)
The Application of Propensity Score Analysis to Non-randomized Medical Device Clinical Studies: A Regulatory Perspective Lilly Yue, Ph.D.* CDRH, FDA,
UNITED NATIONS Shipment Details Report – January 2006.
Document #07-12G 1 RXQ Customer Enrollment Using a Registration Agent Process Flow Diagram (Switch) Customer Supplier Customer authorizes Enrollment.
Document #07-12G 1 RXQ Customer Enrollment Using a Registration Agent Process Flow Diagram (Switch) Customer Supplier Customer authorizes Enrollment.
1 / 18 Network Characteristics of Video Streaming Traffic Ashwin Rao, Yeon-sup Lim *, Chadi Barakat, Arnaud Legout, Don Towsley *, and Walid Dabbous INRIA.
Summary of Convergence Tests for Series and Solved Problems
Jeopardy Q 1 Q 6 Q 11 Q 16 Q 21 Q 2 Q 7 Q 12 Q 17 Q 22 Q 3 Q 8 Q 13
Jeopardy Q 1 Q 6 Q 11 Q 16 Q 21 Q 2 Q 7 Q 12 Q 17 Q 22 Q 3 Q 8 Q 13
Title Subtitle.
My Alphabet Book abcdefghijklm nopqrstuvwxyz.
Multiplying binomials You will have 20 seconds to answer each of the following multiplication problems. If you get hung up, go to the next problem when.
DIVIDING INTEGERS 1. IF THE SIGNS ARE THE SAME THE ANSWER IS POSITIVE 2. IF THE SIGNS ARE DIFFERENT THE ANSWER IS NEGATIVE.
FACTORING ax2 + bx + c Think “unfoil” Work down, Show all steps.
Addition Facts
Overview of Lecture Partitioning Evaluating the Null Hypothesis ANOVA
Lecture 2 ANALYSIS OF VARIANCE: AN INTRODUCTION
The Impact of Soft Resource Allocation on n-tier Application Scalability Qingyang Wang, Simon Malkowski, Yasuhiko Kanemasa, Deepal Jayasinghe, Pengcheng.
Chapter 7 Sampling and Sampling Distributions
Multipath Routing for Video Delivery over Bandwidth-Limited Networks S.-H. Gary Chan Jiancong Chen Department of Computer Science Hong Kong University.
Communicating over the Network
1 Click here to End Presentation Software: Installation and Updates Internet Download CD release NACIS Updates.
1 Implementing Internet Web Sites in Counseling and Career Development James P. Sampson, Jr. Florida State University Copyright 2003 by James P. Sampson,
Streaming Video over the Internet
Experimental and Quasiexperimental Designs Chapter 10 Copyright © 2009 Elsevier Canada, a division of Reed Elsevier Canada, Ltd.
Biostatistics Unit 5 Samples Needs to be completed. 12/24/13.
Table 12.1: Cash Flows to a Cash and Carry Trading Strategy.
Managing Web server performance with AutoTune agents by Y. Diao, J. L. Hellerstein, S. Parekh, J. P. Bigu Jangwon Han Seongwon Park
1 Quality of Service Issues Network design and security Lecture 12.
ABC Technology Project
AMES-Cloud: A Framework of Adaptive Mobile Video Streaming and Efficient Social Video Sharing in the Clouds 作者:Xiaofei Wang, MinChen, Ted Taekyoung Kwon,
IP Multicast Information management 2 Groep T Leuven – Information department 2/14 Agenda •Why IP Multicast ? •Multicast fundamentals •Intradomain.
VOORBLAD.
1 Developing a Predictive Model for Internet Video Quality-of-Experience Athula Balachandran, Vyas Sekar, Aditya Akella, Srinivasan Seshan, Ion Stoica,
1 RA III - Regional Training Seminar on CLIMAT&CLIMAT TEMP Reporting Buenos Aires, Argentina, 25 – 27 October 2006 Status of observing programmes in RA.
Factor P 16 8(8-5ab) 4(d² + 4) 3rs(2r – s) 15cd(1 + 2cd) 8(4a² + 3b²)
Squares and Square Root WALK. Solve each problem REVIEW:
1..
© 2012 National Heart Foundation of Australia. Slide 2.
Lets play bingo!!. Calculate: MEAN Calculate: MEDIAN
Chapter 5 Test Review Sections 5-1 through 5-4.
GG Consulting, LLC I-SUITE. Source: TEA SHARS Frequently asked questions 2.
1 Using Bayesian Network for combining classifiers Leonardo Nogueira Matos Departamento de Computação Universidade Federal de Sergipe.
Addition 1’s to 20.
25 seconds left…...
Week 1.
We will resume in: 25 Minutes.
©Brooks/Cole, 2001 Chapter 12 Derived Types-- Enumerated, Structure and Union.
A SMALL TRUTH TO MAKE LIFE 100%
PSSA Preparation.
McGraw-Hill©The McGraw-Hill Companies, Inc., 2001 Chapter 16 Integrated Services Digital Network (ISDN)
Basics of Statistical Estimation
Compiling Path Queries in Software-Defined Networks Srinivas Narayana Jennifer Rexford and David Walker Princeton University.
Peter Key, Laurent Massoulie, Don Towsley Infocom 07 presented by Park HoSung 1 Path selection and multipath congestion control.
Presentation transcript:

1 1 Detecting Network Neutrality Violations with Causal Inference Mukarram Bin Tariq, Murtaza Motiwala Nick Feamster, Mostafa Ammar {mtariq, murtaza, feamster,

2 2 Network Neutrality ISPs remain neutral in forwarding traffic irrespective of –Content. voice, video, data –Application. p2p, VoIP, VoD –Participants. Service providers, Google, Hulu, Youtube Not-neutral = Discrimination Focus of this paper –Detecting and Quantifying Discrimination

3 3 Net Neutrality Problem

4 4 Discrimination is Real Glasnost Project:

5 5 Discrimination can take many forms Blocking ports Disrupting connections, e.g., using TCP RST Throttling and prioritizing based on destination or service –Target domains, applications, or content Discriminatory peering –Resist peering with certain content providers

6 6 Problem Statement Identify whether a degradation in a service performance is caused by discrimination by an ISP –Quantify the causal effect Existing techniques detect specific ISP methods –TCP RST (Glasnost) –ToS-bit based de-prioritization (NVLens/NetPolice) –Shaping (ShaperProbe) Goal: Establish a causal relationship in the general case, without assuming anything about the ISPs methods

7 7 Causality: An Analogy from Health Epidemiology: study causal relationships between risk factors and health outcome NANO: infer causal relationship between ISP and service performance

8 8 Does Aspirin Make You Healthy? Sample of patients Positive correlation in health and treatment Can we say that Aspirin causes better health? Confounding Variables: correlate with both cause and outcome variables and confuse the causal inference Aspirin No Aspirin Healthy 40%15% Not Healthy 10%35% Sleep Diet Other Drugs Age Aspirin Health ?

9 9 Does an ISP Cause Service Degradation? Sample of client performances with correlation in ISP and service performance Can we say that Comcast is discriminating? Many confounding variables can confuse the inference Comcast No Comcas t Avg BitTorrent Download Time 5 sec2 sec Client Setup TimeofDay ContentLocation Comcast BT Download Time ?

10 Causation vs. Association (1) Baseline Performance Performance with the ISP Causal Effect = E(Real Download time using Comcast) of Comcast E(Real Download time not using Comcast) G 1, G 0 : Ground-truth values for performance (aka. Counter-factual values) Problem: Generally, we do not observe both ground truth values for the same clients. Consequently, in situ data sets are not sufficient to directly estimate causal effect.

11 Causation vs. Association (2) Observed Baseline Performance Observed Performance with the ISP Association = E(Download time using Comcast) with Comcast E(Download time not using Comcast) We can observe association in an in situ data set. In general,. How to estimate causal effect ( ) ?

12 Estimating the Causal Effect Two common approaches –Random Treatment –Adjusting for Confounding Variables

13 Random Treatment Treat subjects with Aspirin randomly, irrespective of their health Observe new outcome and measure association Association converges to causal effect if confounding variables do not change –Diet, other drugs, etc. should not change Aspirin Treated Not Aspirin Treated = = 0.55 H H H H SS S S HHH H H SSS S

14 Random Treatment (How to apply to the ISP Case?) Ask clients to change their ISP to an arbitrary one –Difficult to achieve on the Internet –Changing ISP is cumbersome for the users –Changing ISP may change other confounding variables, e.g., the ISP network changes.

15 Adjusting for Confounding Variables List confounders e.g., gender ={, } Stratify along confounder variable values and measure association Now association implies causation because there is no other explanation Strata HH H HH H HH H SS S H S S SS HH H H S S S S S HH HHH SS SS Treated Baseline Effect

16 Adjusting for Confounding (How to Apply to the ISP Case?) Challenges –What is the baseline? –What are the confounding variables? –How to collect the data? –Can we infer more than the effect? e.g., the discrimination criteria

17 What is the Baseline? Baseline: service performance when ISP is NOT used –We need to use some ISP for comparison e.g., compare ATT vs. Comcast –What if the one we use is not neutral Solutions: –Use average performance over all other ISPs –Use a lab model –Use service providers mode

18 Determine Confounding Variables (Using Domain Knowledge) Client Side –Client setup (Network Setup, ISP contract) –Application (Browser, BT Client, VoIP client) –Resources (Memory, CPU, Utilization) ISP Related –Not all ISPs are equal; e.g., location. Temporal –Diurnal cycles, transient failures

19 Collecting Data NANO uses a passive client side application –pcap library to infer application performance Throughput, Jitter, TCP loss, TCP Reset packets –Applications associated with traffic –Resource monitoring CPU, Memory, Network utilization Summarized performance stats sent to NANO server for analysis

20 Inferring the Criteria Label data in two classes: –discriminated (-) –non-discriminated (+) Train a decision tree for classification –Rules provide hints about the criteria Criteria: youtube traffic, greater than 1 MB is affected

21 Evaluation Experimental Setup ISP –5 ISPs in Emulab –2 Discriminating Service Providers –PlanetLab nodes –Divided in near (west coast) and far (mid and east coast) nodes –HTTP and BitTorrent service Discrimination –Throttling and discriminatory loss using Click routers

22 Experiments Experiment 1: Simple Discrimination –HTTP Web service –Discriminating ISPs drop packets –Server location is confounding: different ratios of near and far nodes in each ISP Experiment 2: Long Flow Discrimination –Two HTTP servers S 1 and S 2 –Discriminating ISPs throttle traffic for S1 or S2 if the transfer exceeds certain threshold –Server location remains confounding Experiment 3: BitTorrent Discrimination –Discriminating ISP maintains list of preferred peers –Has higher drop rate for BitTorrent traffic to non-preferred peers –Location of servers remains confounding Performance Metric: Throughput in all experiments

23 Result 1: Difficult to identify discriminating ISP without stratification Because location of server affect performance and is also correlated with ISP, it confounds the inference Overall throughput distribution in discriminating and non- discriminating ISPs is similar Simple Discrimination Experiment Long Flow Discrimination Experiment

24 Result 2: Discriminating ISPs are easy to identify with stratification Discriminating ISPs have clearly identifiable causal effect on throughput Neutral ISPs are absolved

25 Result 3: Inferring the discrimination criteria ISP throttles throughput of a flow larger than 13MB or about 10K packets cum_pkts not_discriminated cum_pkts > > discriminated NANO correctly infers the threshold for discrimination

26 Conclusions

27 Sufficiency of Confounding Variables

28 Deployment Status