Five ways ICSI researchers have changed the world

Since 1988, ICSI has brought together scientists from all over the world and across sectors to solve challenging problems in computing and data sciences. Unique collaborations among established ICSI researchers, affiliated UC Berkeley professors, international postdoctoral fellows, and academic, government, and industrial partners have created important research breakthroughs that changed the way that network security and software-defined networking are performed, made artificial intelligence (AI) feasible, increased protections for kids using mobile apps, and created intuitive new ways of interacting with our vehicles, among many other advances. 

Monitoring for network intruders

An important advance in network security monitoring came from work Vern Paxson started while pursuing a PhD at Lawrence Berkeley National Laboratory (LBNL). The work led to a system — now named Zeek — that provides real-time network intruder detection and a rich archive of logs for retrospective security analysis after an internet attack. 

After working on the concept at LBNL for two years, Paxson transitioned the project to ICSI. 

The open-source system publicly debuted in 1998 and grew into a platform used by universities, national laboratories, and supercomputer centers. Robin Sommer, who came to ICSI as a graduate student intern and was a fellow in ICSI’s German postdoctoral visitor program, joined the Zeek project in 2001. Under Sommer’s leadership, the project received major investments from the National Science Foundation (NSF) and U.S. Department of Energy. This allowed Zeek to transition into a system that is heavily relied upon by a wide range of customers, including companies such as Apple, Amazon, Facebook, General Electric, Target, and Visa. 

Sommer credits ICSI for offering a one-of-a-kind collaborative atmosphere, where serendipitous conversations spark endeavors that cross disciplines and bring ideas to fruition. “ICSI is a very unique place,” he says. “You have this pure academic research institute without any teaching obligations. If you have the funding, there is a lot of freedom to pursue your research interests. Even as a young researcher, you are basically trusted to develop your own path — and at ICSI, good things come out of that.”

To provide sustainability, Paxson, Sommer, and Seth Hall founded Corelight as a startup spun out of ICSI in the early 2010s. Corelight has attracted $160 million in venture funding to date and grown into one of today’s leading solutions for network security monitoring. 

A revolution in software defined networking 

Another important network security advance came in the mid-2000s when Martín Casado — a summer intern at ICSI — approached Scott Shenker with an ambitious idea for enhancing network security. Casado was a graduate student working with Nick McKeown at Stanford University at the time. This sparked a collaboration between Casado, Shenker, and McKeown that ultimately led to a new paradigm for Software Defined Networking (SDN). SDN provides a clean modularity for the network control plane, which then allows operators to have programmatic control over their networks. The researchers went on to commercialize the technology through the company Nicira in 2007. The company’s key breakthrough came with network virtualization, which allows individual customers in multi-tenant datacenters to create their own virtual networks. The technology has since become a multi-billion-dollar business for VMWare, which acquired Nicira in 2012. 

Shenker was also an important early contributor to the development of distributed hash tables (DHTs), which were simultaneously developed by four different research groups in 2001. DHTs were the first data structures capable of operating on a global scale. One of those pioneering systems — a content-addressable network, or CAN — was spearheaded by Shenker together with ICSI’s Sylvia Ratnasamy, Paul Francis, Mark Handley, and Richard Karp. Although none of those first DHT systems are used today, their invention eventually led to the NoSQL data stores — flexible database designs that are well suited for unstructured data — that are widely used in modern data centers.

Paving the way to practical AI

Krste Asanovic came to ICSI as a graduate student in 1989, not long after the institute’s founding. Within a few years, he was making groundbreaking advances in machine learning technology — decades ahead of today’s AI renaissance. 

Even though the neural networks being developed in the early 1990s were simple compared to those used today, it was recognized that specialized hardware was necessary to reduce the time and energy needed to perform the enormous volume of computation required for AI. With funding from the Office of Naval Research, the Advanced Research Projects Agency, and the NSF, Asanovic and colleagues designed and successfully fabricated a programmable chip optimized for neural network training. The chip’s architecture incorporated both general and machine-learning-specific vector processing elements on a single silicon die. 

This innovation laid the groundwork for researchers to experiment with ever more complex neural networks. “This was a very big project, spanning neural network applications, software, and hardware,” said Asanovic. “We worked across the whole stack, and the advantage of being at ICSI was that all those folks were there at the same place. It was a concentrated center to do this kind of research.”

Keeping kids safe online

The Children’s Online Privacy Protection Act (COPPA) is a federal law that regulates how online services collect and use information from children under 13. Although this law has existed for over 20 years, Serge Egelman and colleagues at ICSI discovered that not all apps were following the law. 

With funding from the NSF, Department of Homeland Security, and National Security Agency, ICSI privacy research team analyzed around 6,000 kid-oriented mobile app and found critical gaps in COPPA enforcement. Immediately after the study was published in 2018, Google and Apple implemented new rules for children’s apps distributed through their marketplaces and the Federal Trade Commission opened a process to revisit COPPA regulations. 

Primal Wijesekera worked with Egelman on this work and was the primary contributor for the custom Android OS used in the testing. “By looking at close to 6,000 apps, we brought essential data that showed why developers are failing and what could help them to resolve these issues,” he said. “Since our results were pointing to patterns of violations, platforms and regulators could take effective and meaningful actions.”

Egelman said that ICSI was a good place to pursue these projects because the institute provides the freedom to do the work. “Once you get the grant money to pursue a project, you can just focus on doing the project,” he said. “ICSI is free of a lot of the bureaucracy you encounter on a university campus, for example.” In 2019, the Egelman and other researchers involved in the project used their technology as the basis for AppCensus, a startup that builds tools for checking mobile app compliance with a range of relevant privacy and security regulations. 

The researchers say that COPPA noncompliance rates are now drastically reduced, though keeping on top of mobile privacy violations requires continued vigilance and tools to empower both app developers and those responsible for enforcing the rules. “Society needs more places like ICSI where we carry out non-partial research into tools and techniques that touch everyone’s life one way or the other,” said Wijesekera.

New ways to control our vehicles

Spoken commands combined with gestures and eye-pointing are providing a new way to control music and functions such as air conditioning in your car — without taking your hands off the wheel or your eyes off the road. This new way of interacting with our cars is now a reality because of work Wolfgang Wahlster and Nelson Morgan began at ICSI in the 1990s.  

The technology was developed at ICSI in a transatlantic cooperation with the German Research Center for Artificial Intelligence (DFKI). In 2003, the team completed the initial project called SmartKom, which was supported by over €25 million from the German Ministry for Research. In the years that followed, research and engineering teams generated over 200 scientific papers as they translated the technology into consumer-ready interactive tools. One such tool, the SemVox user interface, won first prize in the Car HMI Europe Competition when it launched with the 2017 Audi A8, the first car using the technology to hit the market. 

Wahlster founded DFKI in 1988, where he currently serves as Chief Executive Advisor. “ICSI provides an excellent environment for international cooperation between top researchers in many important fields of computer science as one of the world’s leading research labs,” he said. “Its excellent welcome culture for international visitors combined with the California lifestyle and the Silicon Valley ecosystem make it an ideal place for an exchange with the best postdocs from Germany.”

This story was published in January 2026 as part of a retrospective series highlighting ICSI’s accomplishments and impacts over the years. To learn about our ongoing work, explore our Core Research Themes.