Informatics research that has resulted in impact outside academia in a range of sectors.
Research from our different research themes has led to a range of commercial and social impact. Examples of impact across the various research areas are included below.
Novel fuzz testing approaches for the Android Runtime
Researchers from ICSA have collaborated with ARM to produce a novel binary fuzz testing tool, named DexFuzz. Motivated by their observation that traditional fuzz testing was ineffective on binary-encoded compiler intermediate representations, the researchers invented a suite of techniques to significantly increase the coverage of test cases, and were able to discover over 30 times more bugs than a generic testing tool could in the same time period. DexFuzz was employed by Google to successfully change the specifications for their Android Runtime (ART) system, which underpins 2.5 billion smartphones worldwide.
Code size optimisation for smartphone chipsets
World-leading smartphone chip manufacturer Qualcomm have commercialised a compiler tool produced by researchers in ICSA. In Qualcomm’s flagship modem and WiFi chipsets, the tool reduces firmware code size by up to 12%, enabling the company to enhance the functionality of their chips whilst minimising their size and energy consumption. One of the most significant improvements enabled by the ICSA technology is the addition of 5G capability: to date the compiler tool has been employed in over 175 million 5G handsets worldwide.
NextGenIO persistent memory layer
Researchers in EPCC have collaborated with Intel and Fujitsu to produce a ground-breaking novel architecture for high performance computing (HPC) platforms. Named NEXTGenIO, the system can be used for data-intensive applications requiring exceptionally high levels of data throughput and storage. The prototype has been commercialised by Fujitsu in their new HPC server lines, which have formed the basis of supercomputers in multiple organisations worldwide used in applications such as COVID-19 vaccine development, and the prediction of earthquakes, tsunamis and volcanic activity.
Low-power embedded processor design
Work by researchers in ICSA on automating the design of embedded processors resulted in a prototype low-power processor and an ultra-high-speed simulation tool enabling custom processor design by companies. Both technologies were licenced by leading IP developer Synopsys, where they laid the foundations of a long series of processor products. With Synopsys clients including 10 of the world’s top 15 semiconductor companies, 2 billion processors building on Edinburgh research are shipped annually.
Security vulnerability in Fitbit devices
Researchers from ICSA have exposed the vulnerability of Fitbit fitness tracker devices to data manipulation and data theft by hackers. They subsequently made recommendations to Fitbit to improve their device security, prompting Fitbit to develop software patches to enhance the security of their 11 susceptible devices. The ICSA team's work has helped raise awareness of IoT vulnerabilities more generally, and the security recommendations made can be applied to other health-related IoT devices.
Analysis and modelling of laboratory animal behaviour
IANC research into video statistical learning and systems neurobiology has led to demonstrable improvements in the welfare of animal subjects in life sciences research. Commended by the UK Home Office, the technology is already used by international pharmaceutical companies including AstraZeneca, GSK and Bayer, where it has also been shown to yield a reduction in the need for animals.
Speckled computing for improved air pollution and health monitoring
AIAI research into “speckled” computing – miniscule individual computing devices measuring only 1 mm2 – has led to the development of RESpeck, a wearable sensor for monitoring a patient’s breathing rate, and AirSPECK, a self-sustaining pollution-measuring sensor. Used together, these tiny devices can estimate the number of damaging airborne particles in the environment, and have been used in studies on the effects of air pollution in Delhi. They have also been piloted by the NHS in for the early detection of chronic lung disease.
Provable proof-of-stake protocol for new blockchain platform and cryptocurrency
Theoretical LFCS research into distributed ledger technology has led to Ouroboros, the first ever provably secure proof-of-stake protocol for use in blockchains. The work has been adopted by technology company IOHK to underpin the blockchain platform Cardano and associated cryptocurrency Cardano ADA, which has attracted more than 80,000 users and reached a market capitalisation of in excess of US$ 30 billion. Ouroboros has been further applied in the delivery public and economic services, with IOHK announcing partnerships with the national governments of Georgia, Ethiopia, and Mongolia.
Graph databases strengthened by formal semantics for graph query languages
Research in LFCS has produced a formal semantics for the graph query language (GQL) Cypher which underpins the popular database management system Neo4j. The formal semantics produced by the LFCS team removed the possibility of ambiguities in implementations of Cypher that could lead to inconsistent query results. The research has led to the provision of stronger Cypher-driven products for Neo4j’s owner, Neo4j Inc, and its clients, who number thousands of organisations ranging from startups to Fortune 500 companies. The work has also informed the ISO decision to approve the creation of a GQL standard, ultimately setting the course for the wider relational database industry.
Effect handlers for improved software infrastructure
Researchers from LFCS have invented the Effect Handler programming construct and proved it to be a versatile, highly expressive programming abstraction. Effect Handlers have been adopted for software infrastructure by companies worldwide – including GitHub, Uber and Facebook – and have resulted in improved developer productivity and increased product performance for such businesses.
Researchers from IPAB have developed new approaches to dense visual mapping, shared autonomy and motion planning in robots. Partly funded by Hitachi, the company has since used the research to develop an integrated intelligent warehousing system in which a picking robot and automated guided vehicle can together pick items 40% faster than before.
Computer science education
Collaborative research between ILCC and the Universities of Glasgow and Abertay has brought about significant change in the Scottish school curriculum as it relates to computer science. Teaching resources stemming from the research have reached over 650,000 pupils in 2,500 schools. In addition to the bring new skill sets to teachers and in turn their students, the research has informed public policy in Scotland more widely and begun to close the UK digital skills gap.
A toolkit for neural machine translation
Researchers from ILCC have developed Marian, a pioneering toolkit for neural machine translation (NMT). Marian incorporates a number of the team’s innovations in the field, including their influential work on byte pair encoding and back-translation techniques which together enhance the ability of NMT systems to include niche vocabulary and unusual words. Marian has been open source since 2016, and has been used by companies such as Microsoft and Lingo24, as well as laying the foundation for WIPO Translate, the software used by the UN’s WIPO agency for verifying patents.
Neural networks for speech-to-text transcription
ILCC researchers have developed novel neural network algorithms for automatic speech recognition, which have enabled speech recognition capabilities not previously available. Commercialised by Emotech and Quorate, the technology has resulted in a new AI assistance robot; improved workflows for the editors of Hansard, the UK’s official parliamentary record; and enabled Red Bee Media to improve the efficiency of their provision of subtitling services to customers including the BBC and Channel 4.
Speech-driven animation for “talking heads”
Research in ILCC into speech-driven animation has resulted in a novel technology that enables lifelike computer-generated “talking heads” to be created using only an audio recording. The technology has been commercialised by Edinburgh spinout Speech Graphics, now a global leader in speech animation services to the video games industry. Boasting nine out of the world’s top 10 games publishers as customers, the company has brought never-before-seen animation features to titles enjoyed by more than 400 million gamers, which have been lauded by games critics and professional gamers alike.
More detail of each of our research themes can be found on each of the theme pages.
Data Science and Artificial Intelligence