Professor Nishanth Sastry

My work focuses on the structures and architectures used for disseminating and consuming content online. This encompasses a range of topics both at the infrastructure (computer networks) and people (social networks) levels. My long-standing approach has been to apply the lens of data analysis to large, real-world data sets of social networks and computer networks, and using the patterns gleaned therein to design and develop better systems and architectures that are "fit-for-purpose" in deployed or deployable systems. This highly empirical approach, which may be termed "Internet Data Science", has allowed me to work with partners such as BBC R&D, Vodafone R&D and Cisco Systems, as well as drive social and public policy impact, with public bodies and organisations such as the UK Parliament, Wikimedia Foundation and the Samaritans:

Peer-Assisted Content Delivery and Edge Networking Architectures

• Using anonymised data from the equivalent of half the UK population accessing BBC iPlayer, a series of papers showed how the load on UK's networks can be decreased by more than half, and how content delivery can be made greener and more sustainable. We also proposed architectures and theoretical evaluations for peer-assisted offloads and cache-aided D2D, based on individual preferences.
• Using ~3TB of data from Facebook Live, we identified a number of opportunities for edge caching and city-level data centres to decrease and cellular data usage for users.

Computational Politics, Partisan News and Filter Bubbles

• We have worked with Facebook, and scholar/activists to quantify biased representations in news and social media during the Russia/Ukraine conflict.
• We contributed to a massive exposé by Buzzfeed News on the effect of hyper-partisan news and filter bubbles in the 2016 US Presidential elections. 
• Working with the UK House of Commons Library and the Wikimedia Foundation, we have identified patterns of digital citizen engagement on Twitter and Wikipedia, with an "anti filter-bubble" of significant cross-party conversations between Members of Parliament and citizens.

Web tracking and Privacy

• We have deployed a Chrome extension to understand web tracking which is being used by several thousand users across the globe. Using this, we have launched possibly the largest in-the-wild multi-country study of trackers, developed the 'tangle factor' metric to quantify and compare different efforts (e.g., ad blockers) to improve privacy, and to understand how much GDPR-enforced cookie consents have changed the numbers of third party cookies for real users (Answer: not much!)
• With Telefonica Research, we have developed techniques to understand differential tracking of different demographics, showing that right-leaning partisan websites in the USA track more than left-leaning websites 

Harnessing Crowdwork and Cross-Social Network Phenomena

• We have identified several phenomena relating to how users manage their different identities across social platforms. Social bootstrapping showed how social logins help grow nascent social networks such as Pinterest. With Max-Planck Institute, we looked at how to transfer trust across platforms. With Penn State, we identified patterns in how users adapt their profiles to different platforms based on whether they capture formal settings (e.g., LinkedIn) or informal friendships (e.g., Facebook, Instagram).
• We have also developed methods to harness the work of crowds. Predicting Pinterest showed how individual activities on social networks can be treated as a distributed human computation and automated based on crowd work. FaceLift, a collaboration with Bell Labs Cambridge, developed mechanisms to propagate binary rankings of beauty over pairs of images to a whole corpus, and developed a Generative Adversarial Network (GAN) to beautify urban scenes. 

In all the above, "we" refers to me and the group of talented hardworking students and postdocs with whom I collaborate. If this sort of research sounds interesting, and you would like to work with me, feel free to send me an email, mentioning "life=42" to indicate you have read through this text.