CoNEXT '13: Main Street, Wall Street (Session 1)

Crowd-assisted Search for Price Discrimination in E-Commerce: First results Presenter: Nikolaos Laoutaris (Telefonica Research) Story behind: You check a hotel price online, and observe that the quoted price is different than the one given by the same website, for the same product, at the same time to a different user (e.g. a user in a different country) → price discrimination (situation where two consumers are charged differently for the same product; based on how much they are willing to pay) Rumoured to be a problem in e-commerce since the provider has a lot of information available that gives them clues about customer’s behavior (e.g. shopping history, geographic location, behavior on website) Earlier study showed that they indeed observed different prices based on location, using Planetlab nodes as their measurement points In this work: scaled the measurement study, primarily focussing on crowdsourcing Contribution: $heriff, a browser plugin/extension that allows a user to check a price for an item he/she is interested in, and examines differences in prices given to other users 340 beta users, 20 retailers with price variations, monitored 100 products from each retailer Determined the price ratio (max/min price) for each retailer and found that it is larger than 1 for every retailer; for some retailers it is up to 2.0 (i.e. some users pay double the price than other users) Country pairwise comparison: some countries are equally expensive (e.g. Germany vs. Spain), whereas some other countries are cheaper across the board (e.g. Brazil) Determined pricing policies: observed multiplicative rules (e.g. a user from a particular country always pays 30% more a product than a user in another country); multiplicative + additive rules (e.g. pay 30% more + $30 extra for all items in one country) Check out the extension at: pdexperiment.cba.upc.edu Q: What’s the big surprise here? A: No surprise. Our goal was to quantify the extent of price discrimination in e-commerce. Before there were only rumours that price discrimination is happening. Q: How much discrimination can be attributed to factors like differences in taxation across countries? A: For all results, we make sure that we use the initial price before added tax for our comparisons. Q: Is the next step the setup of a brokering system, such that as a customer you can get the best price? A: We thought about it, but it is a complex endeavour. For example, many retailers do not allow shopping from one location yet shipping to another location. However, this might be possible to solve using proxy mechanisms, e.g. for shipping companies like Borderlinx. Q: Some price discrimination can be caused by elasticity in pricing, i.e. you would observe different prices even for a single user? A: Correct. We have some anecdotal evidence that companies have large number of employees just focusing on setting the right prices.

A first look at IPv4 transfer markets

Presenter: Ioana Livadariu (Simula Research Laboratory)

• Situation: Internet registries run out of free IPv4 address space, and IPv6 is only adopted slowly; three RIRs made transfer markets legal
• Approach: analysis of published transfers (list published by RIRs), and detect transfers in the wild to determine whether some transfers are not announced (using BGP routing tables)
• Observed increasing number of published transfers
• Legacy allocation accounts for 40% of all address space; 75% of published transfers are from legacy allocations → healthy redistribution
• Question asked: Are transferred addresses actually used or are they merely hoarded?
• Observation: 85% of transferred blocks are routed after transfer
• Buyers need addresses more than sellers; higher utilization of non-transferred blocks (sellers use between 0.9 and 5.3% of their allocation, buyers use 5 - 19% of their allocation)
• Determined that IPv6 deployment is not expected to eliminate the need for IPv4 addresses (48% of buyers went to IPv4 market before deploying IPv6)
• BGP data analysis resulted in candidate list of possible transfers (prefix for which there is a change in the origin AS); designed filters to remove candidates which observed changes unrelated to transfers
• Observed order of magnitude more candidate transfers compared to published ones (investigating manually)

Q: Is the market increasing?

A: Yes. We see an increasing number of candidate and published transfers.