Application-Specific Secure Gathering of Consumer Preferences and - - PowerPoint PPT Presentation

application specific secure gathering of consumer
SMART_READER_LITE
LIVE PREVIEW

Application-Specific Secure Gathering of Consumer Preferences and - - PowerPoint PPT Presentation

Application-Specific Secure Gathering of Consumer Preferences and Feedback in Information-Centric Networks Reza Tourani, Satyajayant (Jay) Misra, Travis Mick Computer Science Department New Mexico State University New Mexico State University,


slide-1
SLIDE 1

Application-Specific Secure Gathering of Consumer Preferences and Feedback in Information-Centric Networks

Computer Science Department New Mexico State University

Reza Tourani, Satyajayant (Jay) Misra, Travis Mick

New Mexico State University, NM

slide-2
SLIDE 2

Outline

v Introduction and Motivation v Problem Definition v Requirements and Preliminaries v Feedback Collection and Delivery Approaches v Conclusions and Future Work

New Mexico State University, NM

slide-3
SLIDE 3

Outline

v Introduction and Motivation v Problem Definition v Requirements and Preliminaries v Feedback Collection and Delivery Approaches v Conclusions and Future Work

New Mexico State University, NM

slide-4
SLIDE 4

New Mexico State University, NM

Client mining is widespread.

slide-5
SLIDE 5

New Mexico State University, NM

Benefits of client mining and recommender systems.

Influence on 80% of hours streamed at Netflix (2016) Approximately 35% increase in Amazon revenue (2013) 50% of LinkedIn job applications and job views by members (2011)

slide-6
SLIDE 6

New Mexico State University, NM

Amazon Cloud CDNs Netflix Data Center (www.netflix.com) Client

New user registration User account billing

Netflix Communication Flow

Redirecting users

Netflix Server

Authentication Manifest File Periodic Updates CDN Routing DRM

slide-7
SLIDE 7

New Mexico State University, NM

www.hulu.com

(Hosted by Akamai)

Client

Client-Server Interaction in Hulu

s.hulu.com

(Hosted by Akamai)

t.hulu.com

(Hosted by Huku)

3 CDNs

slide-8
SLIDE 8

Outline

v Introduction and Motivation v Problem Definition v Requirements and Preliminaries v Feedback Collection and Delivery Approaches v Conclusions and Future Work

New Mexico State University, NM

slide-9
SLIDE 9

Request flow in ICN multi-level architecture.

New Mexico State University, NM

slide-10
SLIDE 10

New Mexico State University, NM

Data flow in ICN multi-level architecture.

slide-11
SLIDE 11

Pervasive caching eliminates contacting provider for popular content.

New Mexico State University, NM

Cache Hit!

slide-12
SLIDE 12

How to track client without communication?

New Mexico State University, NM

Caching undermines gathering

  • f access statistics.
slide-13
SLIDE 13

Outline

v Introduction and Motivation v Problem Definition v Requirements and Preliminaries v Feedback Collection and Delivery Approaches v Conclusions and Future Work

New Mexico State University, NM

slide-14
SLIDE 14

New Mexico State University, NM

ICN requirements for successful client mining.

Secure feedback collection

Content Provider Independent Preserving User Privacy Precise Statistics

slide-15
SLIDE 15

New Mexico State University, NM

Content Categorization

Generated Beforehand Publicly Available Require Access Control Generated On-Demand

Static Content Dynamic Content Private Content Public Content

  • Generated in advance
  • Require access control
  • Cacheable
  • Generated in advance
  • Available publicly
  • Cacheable
  • Generated by request
  • Available publicly
  • Generated by request
  • Require access control
  • Non-Cacheable
slide-16
SLIDE 16

New Mexico State University, NM

Static-Public is the largest content category.

66% 34%

Content Type in North America

Static-Public Other

slide-17
SLIDE 17

64% 36%

Mobile Access Traffic in North America 2016

Encrypted Un-Encrypted

New Mexico State University, NM

A bigger portion of mobile access traffic is encrypted in comparison to fixed access traffic.

29% 71%

Fixed Access Traffic in North America 2015

Encrypted Un-Encrypted

37% 63%

Fixed Access Traffic in North America 2016

Encrypted Un-Encrypted

Spotlight: Encrypted Internet traffic (https://www.sandvine.com/trends/global-internet-phenomena)

slide-18
SLIDE 18

Outline

v Introduction and Motivation v Problem Definition v Requirements and Preliminaries v Feedback Collection and Delivery Approaches v Conclusions and Future Work

New Mexico State University, NM

slide-19
SLIDE 19

Feedback Collection and Delivery

New Mexico State University, NM

Preference Tracking Mechanisms

Manifest-Free

Collection by Intermediate routers Collection by Clients Collection by ISP's Server

Manifest-Based

Manifest from Provider Manifest from ISP's Server

slide-20
SLIDE 20

Drawbacks of Collection by Intermediate Routers

New Mexico State University, NM

Collection Event

Per-Interest Per-Hit

Drawbacks

Redundant Statistics Coarse-level Statistics Computation Overhead Lack of Client ID

Drawbacks

Coarse-level Statistics Computation Overhead Lack of Client ID

slide-21
SLIDE 21

Drawbacks of Collection by Clients

New Mexico State University, NM

Approaches

Content Partitioning Access Control Enforcement

Drawbacks

Unknown Partition Size Partition Publication Dependency on Online Server Communication Overhead

Drawbacks

Suitable for Private Content Dependency on Online Server

Communication Overhead

slide-22
SLIDE 22

Collection by the ISP’s Designated Server

New Mexico State University, NM

Benefits

Reduced Latency Cache Utilization

Independent of Provider

Drawbacks

ISP-Provider Interaction Inaccurate Statistics

Provider

  • Offload

Decryption Key

  • r Content

Partition ISP's Server

  • Stores Statistics
  • Returns

Requested Key or Content User

  • Request

Decryption Key

  • r Content

Partition

slide-23
SLIDE 23

Manifest-Based Approaches

New Mexico State University, NM

Manifest Delivery

Provider

(Un-cacheable)

ISP’s Server

(Cacheable)

Drawbacks

Extra Latency Provider’s Availability Un-cached Content

Drawbacks

Single Point of Failure Network Bottleneck

slide-24
SLIDE 24

Outline

v Introduction and Motivation v Problem Definition v Requirements and Preliminaries v Feedback Collection and Delivery Approaches v Conclusions and Future Work

New Mexico State University, NM

slide-25
SLIDE 25

New Mexico State University, NM

How about evaluation?

v Manifest-based approaches scale much better than the other schemes. v Communication overhead as means of evaluating efficacy of the approaches – Manifest based approaches scale better. v Manifest-based approaches introduce fixed amount

  • f
  • verhead per content and the amortized cost will be low.

v There is a theoretical upper bound

  • n

the required communication overhead per content: Overhead = 𝐸𝑗𝑡𝑢𝑏𝑜𝑑𝑓𝑑𝑚𝑗𝑓𝑜𝑢 − 𝑡𝑓𝑠𝑤𝑓𝑠 ✕ 𝑁𝑏𝑜𝑗𝑔𝑓𝑡𝑢𝑡𝑗𝑨𝑓

slide-26
SLIDE 26

v Direct interaction between client and provider with/without help of routers inaccurate and non- scalable. v A viable feedback collection mechanism should leverage caching. v Manifest-based feedback collection approaches are more scalable, especially if it involves infrastructure at the ISP. v Comprehensive evaluation

  • f

manifest based approaches (Provider vs. ISP server) and identify which approach in the other class comes closer.

New Mexico State University, NM

Conclusions and Future Work

slide-27
SLIDE 27

Thank you!

Email:misra@cs.nmsu.edu

New Mexico State University, NM

Research funded by the US National Science Foundation and the US Dept. of Defense.