1. ITU Workshop on "Future Trust and Knowledge Infrastructure", Phase 1 Geneva, Switzerland, 24 April 2015 Tussles for Edge Network Caching Patrick Poullie Jr. Researcher Communication Systems Group University of Zurich, poullie@ifi.uzh.ch

2. © 2015 UZH, CSG@IFI Contents  Investigated scenario  Tussle Analysis  Applied Tussle Analysis –Stakeholders and tussles for investigated scenario –Evolution of one tussle  Conclusions Slide 2 of 10

3. © 2015 UZH, CSG@IFI Investigated Scenario  Caching and prefetching of social content in uNaDas –Aspect of fog computing  uNaDa: user-controlled Nano Datacenter –E.g., home router with custom software and storage –Under control of end-user  Social content –Content from social networks (e.g., Facebook) –More complex in terms of caching compared to pure multi- media platform content  Goal: traffic reduction and QoS increase –Content served directly from edge network

4. © 2015 UZH, CSG@IFI Home routers Investigated Scenario Edge ISP End-users Internet Social Networks Slide 3 of 10 Traditional data delivery

5. © 2015 UZH, CSG@IFI Nano data centers Investigated Scenario Edge ISP End-users Internet Social Networks Slide 3 of 10 Envisioned approach

6. © 2015 UZH, CSG@IFI Tussle Analysis  Y.3013: Socio-economic Assessment of Future Networks by Tussle Analysis  Complementation of future network design and standardization by a socio-economic assessment –Analysis goal is to find satisfying outcome for all stakeholders  Developed in SG13 – WP3 – Question 16 –Environmental and socio-economic sustainability in future networks and early realization of future networks Slide 4 of 10

7. © 2015 UZH, CSG@IFI Tussle Analysis’ Three Steps Step 1: Identify all affected stakeholders Step 3: Identify means stakeholders will take according to their interests and resulting interactions Step 2: Identify stakeholders’ interests and conflicts between these and means available to them tussle Tussle cannot be avoided by technology changes No Stakeholder can or wants to change the outcome Slide 5 of 10

8. © 2015 UZH, CSG@IFI Scenario Stakeholders and Interests  End-users –High QoS, privacy, energy and bandwidth utilization  Internet Service Providers (ISP) –Reduce network load (particularly at peering points), sell in-network caching  Cloud Service Providers (CSP) –Content control, data mining, and ad placement  Third parties / data-miners / malicious users –Want access to social content

9. © 2015 UZH, CSG@IFI Scenario Stakeholders and Interests End- users Internet –High QoS –Privacy –energy and bandwidth utilization Internet Service Providers –Reduce network load –Sell in-network caching ISP Cloud Service Provider (CSP) –Content control –Data mining –Ad placement Malicious users / data miners –Access to social content Slide 6 of 10

10. © 2015 UZH, CSG@IFI Scenario Tussles  Content control –CSPs not want third parties to mine their data  End-user privacy –End-users not want third parties to access their private info  uNaDa deployment –End-users require compensation for hosting uNaDa  Change in peering traffic –uNaDas may serve users in other access networks –Increases traffic for ISPs with uNaDas!  Distribution of illegal content Slide 7 of 10

11. © 2015 UZH, CSG@IFI Tussle Evolution (1)  uNaDas get deployed and store content unencrypted –High traffic savings for CSPs and ISPs –Low content control for CSPs  End-users protest due to privacy loss and demand –Thus, uNaDas only store content accessible to owner Comprehensible privacy concept (RFC 7234 not applicable) –Decreases caching efficiency  CSPs distribute encrypted content to uNaDas and handle keys centralized –Regain content control –User privacy ensured –High caching efficiency Slide 8 of 10

12. © 2015 UZH, CSG@IFI Tussle Evolution (2) Tussle: Content Control uNaDas cache only content that is accessible to its owner Tussle: End-user Privacy CSPs have sole control over content Content is served by traditional data centers Regulator intervenes End-users feel their privacy is protected 3. Parties can access content without permission End-users All content cached unencrypted CSP Malicious users / data miners uNaDs store encrypted content, keys are provided by servers End-users feel their privacy is not protected More control but less caching efficiency Slide 9 of 10

13. © 2015 UZH, CSG@IFI Conclusions  CSP need to support uNaDas –Otherwise content cannot be cached  Access providers need to support uNaDas –Else end-users can not be compensated for installing them –Open tussle: unwanted peering  Promising approach –Caching content encrypted and handling keys centralized –One set of keys for each user and privacy level Slide 10 of 10

14. © 2015 UZH, CSG@IFI Acknowledgements