Section: Research Program

User Centric Networking

Billions of people are using the Inernet with different levels of satisfaction concerning the performance. This means that the past research challenges such as efficiency and scalability of Internet protocols are no longer perceived by the users as important challenges to address. However, as the number of Internet devices and bandwidth requirements continue to explode, they still need to be addressed and represent mandatory properties of any new protocol proposed. Apart those well-known research problems, new research challenges are appearing on the design of services centered on the user needs. Thus, we envision a shift from network-centric research challenges to user-centric networking research challenges.

The main consequence of the Internet complexity for the users is the opacity. Users do not have any way to understand and control what their Internet services are doing, which clearly violate their citizen rights. We are witnessing currently a strong interest of Internet providers in this domain along with a growing number of projects funded by service providers (such as Google or Microsoft) on transparency. We define in the following two different rights that we consider as the ones to be addressed for the next decade. DIANA will articulate its research effort around these two rights.

• Service transparency

  The first consequence of Internet complexity is its opacity. The second consequence is the non-predictability of the quality it offers to end-users. It is fundamental for users to be aware of what is going-on on their Internet access and to evaluate the quality they are experiencing, or they can expect, in terms of the different applications they run. For that, some fundamental questions must be answered: What quality can I expect from my Internet access? Why is my service not running properly? Are there any private data sent on the Internet and where? Whereas these questions could be answered with classical measurement techniques (such as ping, traceroute, and tcpdump) in the past, they are vastly more complex today. Indeed, nowadays, mobile device providers, applications/services designers and mobile operators all have conflicting interests and no incentive for real transparency. Regular measurement techniques are either blocked (e.g., traceroutes are blocked by mobile providers to prevent topology discovery which is considered as an industrial secret) or impossible (e.g., tcpdump cannot run on mobile devices because the necessary APIs are not exposed to the application programmer, or simply not implemented in the device drivers). As a consequence, new dedicated measurement platforms are required to work around the existing limitations, and in particular (i) to reveal the reality of the network behind our device and the services we connect to, (ii) to shed light on the quality we can expect from our access in terms of the different applications we run, and (iii) in case of a problem, to help the end user diagnosing its root causes. For instance, to diagnose privacy leaks, we need to perform OS instrumentation and build dedicated experimental platforms to break SSL encryption widely used by services to hide their functioning.

• Open content access, sharing and control

  Users must, at any time, keep the control on their content, that is, seamlessly retrieve them and control who can access them. Today proprietary solutions, such as Google Drive or iCloud, partially solve the problem of accessing content seamlessly on heterogeneous devices, but at the cost of losing control on them. However, several important questions must be answered in this context: where my data is localized physically, who has the right to access my confidential documents, who has actually accessed my photos, how can I be sure that this document is permanently deleted? Having an open access (i.e., independent of a specific vendor) and the possibility to control who is accessing content must be a fundamental right. This means that whatever the device, the operating system, the location, and the available Internet provider, users have the right to seamlessly and efficiently access their content without losing control on them. Protecting users from service misbehaviors and privacy leaks is a difficult task because it requires sophisticated and deployable architectural modifications. One example is the case of a poor video streaming quality. Whereas video streaming is today the most popular application (in terms of aggregate traffic) in the Internet, it is hard for a user to diagnose and solve issues because Internet actors (ISPs and service providers) have conflicting interests and no reason to collaborate. A vivid example is the one of Free, a French ISP, which is limiting the throughput for YouTube streams in order to put pressure on Google to compensate the ISP investments. This is clearly against the citizen rights, and it requires working around the throughput limitations using a dedicated and open indirection infrastructure. Another example is the one of the free application market. Most of free applications embed advertisements that are operated by third parties. Whereas most applications do not need to send data on the Internet, private information are leaked in order to run targeted advertisements only. This issue cannot be solved with simple ad-blocking on the device, because the operating system of mobile devices does not allow interposing on applications behavior. Also, blindly blocking all ads might challenge the market of free applications. The only solution is to build a dedicated open infrastructure that filters out private information while still making the business of targeted advertisement possible.