(We created the Comcast Innovation Fund to support important research into the future of the Internet, with a focus on broadband, security and open-source development. In this monthly series, we highlight grantees and their work.)

Carpenter and Dittrich: Privacy in Open Source

The Comcast Innovation Fund is pleased to announce the award of a grant to two independent researchers, Katherine Carpenter and Dave Dittrich to support research into best practices for using open source tools while protecting private information. The current research builds on work that was funded by the Department of Homeland Security.

These best current practices will help development teams to manage open source code development that involves access control mechanisms, such as passwords, certificates, and cryptographic keys, that must be kept secret, while the source code and system configuration files that use these secrets are themselves made public.

A key goal of the work is to model general best practices that can work across a variety of tool sets. They hope to create documentation in the form of a paper or short e-book and working source code examples that help software developers follow a best-practice model. This could serve as in-house training material for developers.

Documentation will cover representative source code workflow processes that present issues with management of secrets, guidance on the process for recovering in cases in which private information is exposed, and several models for reducing risk. This research could form the basis for future presentations and/or hands-on tutorials at OSS and other conferences.

UCSD/CAIDA: Advancing Measurement With PacketLab

Another Innovation Fund grant went to an exciting new collaboration at the University of California San Diego, between the Computer Science and Engineering (CSE) Department and the Center for Applied Internet Data Analysis (CAIDA). The grant will support a new approach to developing and maintaining measurement infrastructure by defining a universal endpoint measurement interface called PacketLab. This research program is overseen by Research Scientists Kirill Levchenko (CSE) and Research Scientist Amogh Dhamdhere (CAIDA), with support from other researchers at CAIDA.

PacketLab is built on two key ideas: It moves the measurement logic out of the endpoint to a separate experiment control server, making each endpoint a light-weight packet source/sink. At the same time, it provides a way to delegate access to measurement endpoints while retaining fine-grained control over how one's endpoints are used by others, allowing research groups to share measurement infrastructure with each other with very little overhead.

By making the endpoint interface simple, UCSD/CAIDA also aims to make it easier to deploy measurement endpoints on any device anywhere, for any period of time the owner chooses. As part of this work, they will develop an endpoint agent that can be installed on new or existing measurement endpoints. The agent will provide researchers and operators a universal interface for conducting network measurements.

UCSD/CAIDA hopes that PacketLab can be a measurement interface that can accommodate the research community's demand for future global-scale Internet measurement. The project is intended to be community-driven, and one of its main outcomes will be community support for a universal measurement endpoint interface across existing measurement platforms. The design document and code will be available to the public under an open source license. Results of this work will also eventually be submitted for publication to top networking and measurement conferences.

Internet Systems Consortium: Open Source DCHP Software

The Internet Systems Consortium (ISC) is using its Innovation Fund grant for continued development of its open source Kea DHCP software.

Kea is a modern open source DHCPv4 and DHCPv6 server that runs on UNIX and LINUX systems. This new work supported development of Kea version 1.3, featuring support for shared subnets, which is the feature most frequently mentioned on the public Kea-users mailing list as a requirement for adopting Kea.

The shared subnets feature enables the server to assign multiple addresses from different IP subnets to devices connected to the same physical network, common in many enterprise networks. This is particularly important for networks that have run out of addresses from the assigned IPv4 subnet and need to add another IPv4 subnet on top of existing one. Similarly, IPv6 deployments that want to modify their addressing scheme may need to support both old and new IPv6 subnets. Furthermore, the popularity of using virtual machines for application deployment is driving new requirements to assign multiple addresses per physical device, and when subnets are used to group related applications together, (e.g. in lieu of a VLAN), that also requires shared subnets.

ISC hopes that by adding this feature, many more people will adopt Kea, making it a more vibrant and widely deployed open source project, as well as help organizations migrate away from their older ISC-DHCP software or other platforms.

The new code is now available as part of the Kea 1.3.0 release.

University of Toronto: Deep Learning for Natural Language Queries

The University of Toronto’s Department of Computer Science received a grant to support research into deep learning for video content retrieval using natural language queries, at the direction of Assistant Professor Sanja Fidler.

The research project will tackle the problem of finding and retrieving videos using complex natural language queries. In particular, they plan to address a scenario where the user describes a movie scene or a news event in natural language, in order to retrieve the relevant clip(s) from a lengthy video.

The problem is very challenging, since one needs to semantically parse videos, and be able to match them to textual queries.  As a result, this research will explore a new way of representing video content in the form of semantic graphs, which may enable much more efficient methods to find relevant video content. If this project is successful, the research will result in development of a new method for querying videos via graphs.

We’re excited to follow the work of all of these researchers, and to try out the new tools and methods they are developing.