Rhodes Scholarship [verify]
Dean’s List (2017-18, 2018-19, 2019-20)
Joseph Lau Luen Hung Charitable Trust Scholarship Hong Kong Award
Tin Ka Ping Scholarship (Innovation & Creativity)
Li Ka Shing Foundation Scholarship for Innovation and Creativity
Hong Kong Unison Scholarship
University Entrance Scholarship
Harmony Scholarship (Home Affairs Bureau, HKSAR Government)
HKSAR Government Scholarship Fund – Reaching Out Award
Br Bernard Guellec Memorial Prize for French
Br Felix Sheehan Memorial Prize for English
ETHGlobal: HackMoney 2020 - Winner (pToken) [1] [2]
CalHacks 5.0 - Winner (Augmented Reality section, Epson Sponsor Award)
SacHacks 2018 - Winner (Smartcar 1st Runnerup, Best First-Time Hack, Best Domain Name, 3rd Overall)
AngelHack 2018 - Champion (Blockchain Category)
Cyberport Creative Micro Fund - Winner
Classified Post Hackathon 2018 - Champion
Li&Fung Hack the Runway 2017 - Champion
HKUST Bizkathon 2019 - 1st Runnerup
CityU Hackathon 2018 - Most Innovative Idea award & 1st Runnerup
EasyShip 24 Hour E-Commerce Hackathon 2017 - 1st Runner up
HardUST 2017 - 1st Runner-up
HKUST-Radica Big Datathon 2017 - 1st Runner up
Access to Justice InnoTech Law Hackathon 2018 - Honorable Mentions
Expand/Collapse items on this page


S. Datta. 2021. Learn2Weight: Weights Transfer Defense against Similar-domain Adversarial Attacks. Under Review In International Conference on Learning Representations
S. Datta, D. Alonso, P. Ferreira. 2021. Fast power spectrum interpolation with ensembling networks. Under Review In Monthly Notices of the Royal Astronomical Society
S. Datta, K. Kollnig. 2021. GreaseDroid: A Paradigm Shift in Mobile App Dark Patterns. Under Review In ACM CHI Conference on Human Factors in Computing Systems Late-Breaking Work
S. Datta. 2021. GIFShop Wizard: Dialogue-based Object-oriented Video-editing. Under Review In ACM CHI Conference on Human Factors in Computing Systems Late-Breaking Work
S. Datta. 2021. Accepted. DeepObfusCode: Source Code Obfuscation Through Sequence-to-Sequence Networks. In Advances in Intelligent Systems and Computing The paper explores a novel methodology in source code obfuscation through the application of text-based recurrent neural network (RNN) encoder-decoder models in ciphertext generation and key generation. Sequence-to-sequence models are incorporated into the model architecture to generate obfuscated code, generate the deobfuscation key, and live execution. Quantitative benchmark comparison to existing obfuscation methods indicate significant improvement in stealth and execution cost for the proposed solution, and experiments regarding the model's properties yield positive results regarding its character variation, dissimilarity to the original codebase, and consistent length of obfuscated code.
S. Datta, C.Z Koval. 2018. Big Data Analysis of Labour Market Dynamics in Freelance Work Platforms In Undergraduates Research Opportunities Program 2017-2018 Proceedings This research aims to explore the underlying relational dynamics between people on freelance work platforms, between employers and employees, and within employees themselves. Particular interest will be placed on gender correlations and the objective to locate the most important variables involved in reasonable pay, as the aim of the study is to identify biases on the platform. Data was scraped from the work platform for all the possible variables available, from tests taken, to work histories, to skills listed, to even general personal descriptions.
A. Goel, K. Slama, S. Datta, L. Dong. 2018. "Predicting Value-based Decision-making using Time Series neural ECoG data" In Data-X 2018 [code] [poster] [data] [library] A project focused on predicting gambling decisions from brainwave electrocorticography (ECoG) data. The motivation behind this project was based on societal implications such as applications to law and regulation of the gambling industry, clinical practice in understanding how neurodegenerative disease or trauma affect decision making, and to some extent even our understanding of free will. The project utilized data recorded by the Knight Lab, a neuroscience lab at UC Berkeley. The ECoG data was collected from the orbitofrontal cortex (OFC) of epileptic patients, who volunteered to participate in research while undergoing intracranial monitoring for surgical planning purposes. Patients were given a task of playing a game in which they could choose either ‘Gamble’ or ‘Safebet’ for 200 trials. Over the course of the semester, we pre-processed, explored, and modeled the data, with the ultimate goal of predicting whether a patient would gamble or not given only the ECoG data, recorded before the patient indicated their decision by a button press. Much exploration was on feature engineering, as there are not many established feature engineering methods in the analysis of ECoG data: examples of features include using the 2.5 and 97.5 percentile values and number of peaks for electrode data in a patient’s trial. Modeling was done through logistic regression, random forest, adaboost, naive bayes, neural nets, and perceptron. Logistic regression fared the best, with one model giving a 10% boost over the baseline accuracy found with Naive Bayes. This result demonstrates that there is information in the OFC neural signal, which can be used to predict decisions before they are made overtly.
S. Datta 2018. HAIV: Computational Precision Medicine against HIV Drug Resistance In HKNGCA Challenge Cup (Finalist) This paper introduces a novel methodology to treat HIV through the combination of existing antiretroviral therapy with mutation prediction algorithm (“HAIV”). The algorithm developed predicts the next possible mutations that a given HIV strain may undergo, and implies doctors can apply the right medication at the right time, hence alleviating the need to apply more variations of medication than needed, further suppressing the possibility for drug resistance. The work is deemed to be novel and without any prior work in the industry thus far, and the applications range from any task requiring the prediction of mutations, or any virus dealing with drug resistance.
S. Datta, S.L. Nam 2017. BreakupBot: An Artificial Intelligence Therapeutic Chatbot Conducting Empathetic Counselling via Heartrate Sentimental Analysis In President's Cup 2016-17 (Finalist) BreakupBot is the title of a chatbot platform that aids people to recover from their romantic breakups. To cater to a heterogeneous audience, there will be several therapy mechanisms in place to ensure every type of lover will be catered to. Some users may wish to think deeply about relationships, while some may need to gain maturity about love in general; in either case, the AI behind the chatbot will personalize itself to the user, and adopt both psychological and philosophical counselling methodologies. Personalization ranges from understanding the user’s lover characteristics, analyzing the user’s stress levels via heartrate, and responding instantaneously to their every query and woe. The technology is adept and fitting for complementing the bot: other than the artificial intelligence system that collects information to share with the user and personalizes it to the user’s taste, the bot features accurate heartrate analysis available to most smartphones via camera-detection. Such a mechanism will aid the bot in empathizing with the user, contrary to other bots who attempt facial recognition or process speech patterns. Moreover, this project merely represents the beginning of further exploration into psychological therapeutic benefits via AI. Once the BreakupBot penetrates the market wide enough to allow entry to other similar products, other therapeutic bots and mechanisms can be designed to streamline human being’s complicated lives and emotions. This project is a goldmine, with respect to the fact that it creates its own category of products: it is not in the personal assistant AI category, but the currently underdeveloped therapy AI category, hence there is a lot of potential for research and profit. At the end of the day, the value of this project is not only its intrinsic research, or its mechanisms, or its cross-departmental field of study; its value comes from robots attempting to solve the problems of human beings at a psychological level.

… And More Papers

Trade Network Analysis of Exports Loss due to Internal Political Events Tested the causal impact of 2019 protests on Hong Kong's export of financial services to partner nations compared to Singapore (the change in flow to multilateral ties relative between the two cities, as a proxy for change in flow between common relationships). Relevant slides: [8-11, 39-40, 42-43]
Using Structural Imbalance to Evaluate Models of International Relations Studied whether international conflicts could be predicted or captured in network imbalances between nodes; we used discrete sign methods to measure structural imbalances for each feature, and also used a continuous reinforcement learning method to calculate weights for each feature in each edge to measure structural imbalance in one graph
Network Analysis of Power and Politics in Academia Iterated through academic network graphs to calculate relative power differences between grad students and their PIs; obtained weights of each power feature in weighted-function of edge weight through gradient descent
Motif detection and clustering of franchise location network graphs Hybrid implementation of motif-detection, bridge-detection, and clustering algorithms to yield sequential coordinates of geographical locations depending on category of product/business, based on network de-anonymization framework
Youtube Sentiment-based Portfolio Selection and Trading Strategy
AIDS Epidemic Simulation with Adjustment for Decision-making from neural ECoG data


Computer Security, Michaelmas Term 2020, Oxford [Week 4] [Week 5] [Week 7] [Special Topics: Adversarial Attacks] [Week 8]
Advanced Security, Hilary Term 2021, Oxford
Networks, Hilary Term 2021, Oxford [Network De-anonymization]
Artificial Intelligence, Hilary Term 2021, Oxford