Precog Diaries – A Roller Coaster Ride

It all began with a mail from PK in November 2014, commending my performance in his courses (CSE 545: Foundations of Computer Security and CSE 51SP: Internet Security and Privacy). It always feels great when a professor pats your back and even more when he offers you to join his research group! Precog is a different place altogether in IIIT Delhi, its not just a Research Lab, it is a “dojo of learning”. Randy Pausch once said “… experience is often the most valuable thing you have to offer …”. I was fortunate enough to spend 18 months here and I believe it has been an extremely rewarding experience. The tribe at Precog is a mix of ‘Super Cool’ PhDs, ‘Fantastic’ RAs, ‘Brilliant’ undergrads and ‘Awesome’ postgrads doing wonderful work on security and privacy in Online Social Networks, all headed by the guru “PK”.

The lab culture is amazingly open and you can approach ANYONE for help and knowledge. It’s great when you have wisdom trees (AKA PhDs) dotted all around the lab and you can seek shelter under any one of them. I personally feel right mentorship is vital for research and Dr. PK has been a very strong guiding force. Initially I worked with Anupama (Ph.D. students) on the problem of social reputation manipulation in OSNs where I learnt a great deal of new technologies and preliminary data analysis. In 2015, Precog was granted a government project where the theme was to understand and analyse disclosure made by employees of an organisation on OSNs.

I am grateful that PK considered to chose me to work on this project which also aligned with my thesis work. We explored six social networks to identify employees of a subject organisation and collected public posts about the organisation. We analysed this data in multiple dimensions like popularity, reach, sentiment, source of influence and geospatiality to discover actionable intelligence and results which could aid the organisation to leverage the information and monitor the content for sensitive disclosure. A system based on this work was developed that can collect, analyse and visualise information being posted by employees of an organisation. I was thoroughly supported by PK and Precog-er’s especially Rishabh (Ph.D. student) sir who collaborated with me on this project. I’d like to extend my gratitude to them for all the guidance and motivation.

Some of the lessons were learnt hard, but I believe that’s important because it makes them so valuable. Apart from work, Precog is also a fun place and we have had loads of celebrations for our achievements in the past year. It has also given me an opportunity to take part in conferences and workshops and interact with researchers having similar interests. It has been a great time at Precog and I will always cherish working here. Some beautiful memories of the past year ..

Below are some pics from the group!

Funfilled and Rich Experience @ Precog

When I started my journey with Precog, I did not had an idea about the fun-filled work culture, cool and real world implicated projects and work done here will motivate me so much.

It was just a year back when we came up with this idea of exploring how images that can create law and order situations spread on social media. I was highly moved that how a single image can spread so easily, gets viral exponentially, reaching new set of people minute by minute, instigating anger resentment, and fanning violence eventually. WOW! All Happened Without Even a Word!!!

And then I said to myself here is a problem that I intend to solve, it was something new to me and perhaps not of much work has been done in this regard…But as a Precog-ger I intend to take it up.

We picked up the problem of quantifying image virality on Twitter by computing the frequency of similar images posted on Twitter. The overall work was divided in multiple phases. Starting with data collection, we collected around 5,000 images from Twitter for pot-boiling incidences. Next step was to compare image features using different methodologies like 3D colour Histogram, DAISY features, ORB and then we came up with Improved ORB technique which gave us the accuracy of around 90%. In the next phase we built a system which is being used by government security agencies. The input to the system is an image and a related keyword and the system quantises the virality based on the frequency of similar images shared. Further to assist security analysts take some proactive preventive measures and understand the spread of an image we also predicted the spread of an image based on its contextual, structural and visual characteristics. Below is a screen shot of the system that I built during my Masters thesis work. You can get the thesis from here, my slides from the thesis here, and a picture (with my evaluators Dr. A V Subramanyam, and PK) that we took when my defense was done!

I would like to owe sincere gratitude to PK for his inputs and giving me the chance to work on such a burning issue, Niharika Sachdeva (Ph. D, IIIT-Delhi), and Dr. Samarth Bharadwaj (IBM-IRL) for guiding me throughout the project.

Apart from work environment I would like to mention about the regular leisure activities that we enjoy. Recently we went to Sariska reserve; the trip was very soothing and relaxing (some pics below). The scenic beauty of the reserve was captivating. I enjoyed observing vivid plants, birds and animals, watching them eating food, making noises… And in all, this trip was one of the few beautiful memories with Precog. Also to mention small celebrations on every success that any member in Precog family gets!

While concluding I admit that I have got insight to the research work and its ethics, gained from vast pool of knowledge and experience that people here possess, firsthand experience of new technologies and real world problem solving ability. I have had amazing time in last one year at Precog! Looking forward for exponential growth here in years to come…