Three more things you need to know about the Responsible Internet

The problem: declining digital autonomy

Societies around the world increasingly depend on digital systems (e.g. algorithms, cloud storage and DNS services) for the well-being of their citizens and for a thriving economy. However, these systems are increasingly developed or run by a small number of economically very powerful companies from the USA or China (e.g. Google, Amazon and Alibaba), which have by now gained disproportionate control over the knowledge, data and technologies on which our modern digital societies depend.

This development has led to a broad consensus in recent years that the digital autonomy of regions that predominantly consume ICT is under pressure, for instance in the Netherlands, Germany and Europe. The concern is that they will have less and less control over how the digital systems designed and operated by ICT producers shape their societies. The risk they foresee is that they slip into 'powerless dependence', which may then lead to them losing control over (or being forced to adjust) their public values, such as tolerance, security, transparency and democracy.

An important part of the problem of declining digital autonomy is that most software and hardware systems are black boxes. As a result, organisations and citizens have no insight into or control over the operation and design of such systems. In fact, they often aren't even aware of how they depend on them. The problematic nature of black-box design has recently been highted by several prominent news stories: a company that supplied network equipment to a Dutch telecom operator was allegedly able to monitor calls; there has been considerable debate about security weaknesses in 5G equipment and internet routers have been (physically) hacked. The problem of declining digital autonomy is particularly strongly felt in the context of emerging safety-critical services (e.g. 5G networks and smart energy grids), which the Dutch government defines as services “so important to Dutch society that failure or disruption will lead to serious social disruption and pose a threat to national security”.

Our proposed solution: The Responsible Internet

The solution we are proposing to address the network communications problem described above is the concept of a Responsible Internet, a security-by-design extension to the Internet’s network infrastructure that allows all stakeholders to rely on the Internet’s operation with a significantly higher degree of confidence and trust than is possible today. A Responsible Internet will accomplish that in two ways. First, by extending the original set of design goals of the Internet protocol suite with three new ones: controllability, accountability and transparency. Controllability implies reliant parties being able to choose the (groups of) operators that they want to entrust with processing and forwarding their data across the Internet, such as operators that have carried out audits to rule out backdoors in their equipment or that fall under a certain jurisdiction. Accountability and transparency imply reliant parties being able to obtain relevant metadata about the Internet’s structure and operation from the network in a verifiable way, thus turning the network infrastructure into a 'glass box' rather than a black box. Together, the three new design goals will complement the traditional security dimensions of confidentiality, integrity and availability. Second, the Responsible Internet will realise the three new design goals through open networks. Those will consist of equipment whose software is open source (e.g. based on the P4 language) and of open hardware modules (e.g. the Open Compute Project). Open networks will allow critical infrastructure operators and other users to (remotely) assess information about the security of network equipment (transparency), which is important as they will typically not know all the network operators that transport their data (the Internet is a network of networks, after all). Also, open networks will allow for experimentation with new network functions on hardware (e.g. the addition of custom headers to IPv6 packets) and non-IP protocols (e.g. SCION in P4).

Property #1: Continued support for freedom of equipment choice

The first property we’d like to expand on is that the Responsible Internet will continue to allow network operators to autonomously decide what equipment they use in their networks. The only requirements are (1) that it provides the functionality for cross-domain controllability, transparency and accountability (e.g. using services, APIs, data models and protocols), and (2) that it is open in terms of software and hardware. A possible evolutionary path is that operators first introduce the functions of the Responsible Internet using conventional closed network equipment and then invest in open programmable hardware to add the extra level of transparency and control. The Responsible Internet will not prescribe the use of network equipment from certain vendors or regions (as in the US’s former Clean Network initiative). In our view, such requirements are unnecessary and are currently also virtually impossible for regions like Europe that don’t have many industry players capable of manufacturing network equipment. In the long term, however, it would also be worthwhile for Europe to build up its own network industry, at least in selected areas (e.g. open networks with remote attestation facilities) to create the more diversified and resilient digital supply chains that the European Commission envisions. As a result of this property, the Responsible Internet can not only be used with the current TCP/IP Internet, but it can also serve as an extension to 'clean slate' networks such as SCION and NDN. In the future, a Responsible Internet may even encompass different co-existing Internet designs.

Property #2: Focus on inter-domain networks with Internet-style architectures

The second key property of the Responsible Internet is that it will focus on 'responsibilising' inter-domain networks with the following architectural characteristics:

• Support for end-to-end communications: networks in which only endpoints determine what data they exchange (using end-to-end encryption), enabling them to introduce new services without the need to modify the network.

• General purpose: networks that can meet the communication needs of a wide range of evolving applications. This is unlike special-purpose networks, which are optimised for specific applications such as voice or sensors.

• Open and accessible: inter-domain networks that enable new networks and new users to connect freely, without permission from a central authority.

• Support for decentralised and distributed ownership and management: networks in which operators are responsible for the equipment under their control and autonomously decide how to organise their infrastructures.

These characteristics correspond broadly to 'the critical properties of the Internet' defined by ISOC, with the difference that we also envision a role for Internet designs that are not based on IP but still align with ISOC’s properties, such as SCION.

Property #3: International set-up

While the objective of the Responsible Internet is to increase the digital autonomy of societies such as the Netherlands or regions such as Europe, we stress that it is not about setting up national, isolated networks, such as the 'Beijing Internet' or the 'Moscow Internet'. Rather, we want a Responsible Internet that is as cross-border as the current Internet. This is also why we switched to emphasising the concept of 'digital autonomy' rather than 'digital sovereignty', because the latter might suggest a focus on nations or specific regions. The international character of the Responsible Internet was discussed in our original paper, but we figured it was important to reiterate it here. :-)

Next steps

We hope that explicitly stating the properties we envisage for the Responsible Internet will position our work more clearly. We’re enthusiastic about starting our CATRIN project soon. We expect this clarification will also help guide discussion at the upcoming TAURIN workshop, where we hope to enable the research community to discuss many novel ideas for getting the Responsible Internet started.

Acknowledgements

This work is part of the 2STiC research programme (Security, Stability and Transparency for inter-network Communications). Website: https://www.2stic.nl/.

SIDN and the University of Twente were partly funded by the European Union’s Horizon 2020 Research and Innovation Programme under Grant Agreement No 830927. Project website: https://www.concordia-h2020.eu/.