1. The SIGACT Community and the NetSE Program: A Match Made in Heaven! Joan Feigenbaum http://www.cs.yale.edu/homes/jf/ Arlington VA ; September 2008

2. “The NetSE program seeks proposals focused on developing new theoretical foundations, principles and methodologies to understand and reason about the dynamics and behavior of current and future large-scale networks, the interdependence among the physical, informational and social networks they embody, and the tradeoffs among communication, computation and storage.” “The emphasis is on creating theoretically grounded architectures that address fundamental policy and design engineering trade-offs, support healthy economic models and promote social benefits.” CISE’s “Cross-Cutting” Solicitation Says: MUSIC TO MY EARS!

3. Theory of Networked Computation (ToNC) NSF Small Grant for Exploratory Research supported two ToNC workshops in 2006. –1 st ToNC Workshop: Nassau Inn in Princeton NJ, February 2006, chaired by Rexford and Feigenbaum –2 nd ToNC Workshop: ICSI in Berkeley CA, March 2006, chaired by Shenker and Feigenbaum Participants articulated a broad, eclectic research agenda that anticipated NetSE goals. Slides, workshop report, and other information at http://www.cs.yale.edu/homes/jf/ToNC.html

4. ToNC (2) ToNC agenda has three major components. –Realizing better networks –Computing on networks –Algorithmic and combinatorial problems created or exacerbated by networks ToNC report advocates collaboration by –Sigact people and Sigcomm people –Theorists and experimentalists Sample ToNC research challenges follow – not an exhaustive list!

5. Complexity Theory of Networked Computation Machine model(s) that capture(s), e.g., massive scale, user self-interest, device heterogeneity, and emergent behavior Resources Reductions Hardness results (for better and for worse) Questions: –“Cook’s Theorem of Networked Computation”? –Computational model ≈ Network architecture?

6. Economic and Strategic Considerations “Ownership, operation, and use by many self- interested, independent parties give the Internet characteristics of an economy as well as those of a computer.” Progress on “algorithmic mechanisms,” including ones for digital-good auctions, combinatorial auctions, load balancing, cost sharing, and routing Questions: –Are “equilibrium” concepts from Econ appropriate? –Are non-monetary exchanges useful? (Think BitTorrent) –Is “irrational” behavior recognizable and manageable?

7. Interdomain Routing IDR: Paradigmatic problem in networked computation –Exemplifies crucial requirements (subnetwork autonomy, subnetwork heterogeneity, massive scale, attack resistance, etc.) –Widely deployed –Hard but unavoidable (and interesting!) Current state of affairs is unsatisfactory. –Wilfong at ToNC Workshop 1: “BGP Horror Story” –Progress over last 10 years (SPP, robustness, economic analysis, etc.) –Still many difficulties  Small changes in policies lead to big changes in behavior.  Hard to realize some natural goals and requirements

8. Interdomain Routing (2) General challenge: Develop a Grand Unified Theory of Interdomain Routing. Some specific goals: –Identify operating regimes in which current approaches work well (ref. Gao-Rexford). –Separate “essential IDR” from BGP. –Clarify relationship to other networked- computational goals. –Transitions and paths to adoption

9. Massive-Dataset Computation Progress over the last 10 years –Useful computational models (e.g., streaming, spot checking, property testing) –Many algorithmic results (especially randomized, approximate, and near-linear) Massive datasets abound in networked computation Grand Challenge: Next-generation search –Adversarial behavior (“google bombing”) –Complex data formats –Personalization: Utility vs. privacy

10. Peer-Produced Information Services Peer production (Benkler 2001) –High modularity –High granularity –Low cost of integration Examples abound in the Internet age. - Blogs, wikis, bulletin boards, etc. –Uncoordinated, “egoistic” development –Commercial successes such as Google, eBay, FaceBook, etc. “leverage peer production.” General challenge: Rigorously define and investigate peer production. –Accuracy/quality of peer-produced information –Computational vs. economic cost of discovery –Computational vs. economic cost of privacy

11. Security of Network Computation Assess security at the network level, not the agent level. Consider quantitative measures of security, not just worst case guarantees. Consider incentives for deployment and adoption. Questions: –After 30+ wildly successful years of security and cryptology research, why is our computing environment so insecure? –Potential users should be able to get meaningful answers when they ask, “How much will it cost me to use this security technology? What will I gain if I use it? What will I lose if I don’t?”

12. Privacy in Networked Computation Robust technological and social trends have led to a dramatic increase in the amount of sensitive information about people and organizations that is created, captured, stored, and traded. Past: “Privacy”  “Confidentiality of info.” Future: “Privacy”  “Appropriate use of info.”? Questions: –New crypto-theory formulations, e.g., “Privacy in public dbs” (Dwork et al.) “Group privacy” (Shmatikov et al.) –Is any meaningful notion of “privacy” compatible with ubiquitous powerful computers and networks?

13. Progress Already! Increasing numbers of papers on NetSE themes at STOC, FOCS, SODA, ICALP, STACS, PODC, DISC, EC, etc. SIGACT researchers have participated in the SING and FIND programs. Questions?