Thread 1: Dataset collection, AI-ML applications, complex audits for environmental sustainability in cities of developing countries. Mainly focussed on measurement and analysis of air pollutant Particulate Matter (PM) and road traffic, in the cities of Delhi and Kolkata. The research work was done by PhD student Sachin Chauhan (2018-), and several MTech students Sagar, Saswat and Chinmay. The work also had significant engineering efforts by some excellent field staffs.
[JCSS24] "FrugalLight: Symmetry-Aware Cyclic Heterogeneous Intersection Control using Deep Reinforcement Learning with Model Compression, Distillation and Domain Knowledge", Sachin Chauhan, Rijurekha Sen in ACM Journal on Computing and Sustainable Societies (ACM Journal), presented in ACM COMPASS 2024 conference |
[NeurIPS23] "Fine-Grained Spatio-Temporal Particulate Matter Dataset From Delhi For ML based Modeling", Sachin Chauhan, Sayan Ranu, Rijurekha Sen, Zeel B. Patel, Nipun Batra in Proceedings of the 37th Conference on Advances in Neural Inforamtion Procesing Systems, (Datasets and Benchmarks Track) (Core A*) |
[COMPASS22_1] "Complexity of Factor Analysis for Particulate Matter (PM) Data: A Measurement Based Case Study in Delhi-NCR", Ismi Abidi, Sagar Ravi Gaddam, Saswat Kumar Pujari, Chinmay Shirish Degwekar, Rijurekha Sen in Proceedings of ACM SIGCAS and SIGCHI Computing and Sustainable Societies (Core unranked) |
[NeurIPS20] "EcoLight: Intersection Control in Developing Regions Under Extreme Budget and Network Constraints", Sachin Chauhan, Kashish Bansal, Rijurekha Sen in Proceedings of the 33rd Conference on Advances in Neural Inforamtion Procesing Systems (Core A*) |
[ICTD19] "Embedded CNN based vehicle classification and counting in non-laned road traffic", Mayank Singh Chauhan, Arshdeep Singh, Mansi Khemka, Arneish Prateek, Rijurekha Sen in Proceedings of the 11th International Conference on Information and Communication Technologies and Development (Core C) |
Thread 2: Security and privacy of running data intensive applications on Edge/IoT platforms. As broadband network connectivity is not ubiquitous in developing countries, sending sensor data from the field to the cloud is not always feasible. Thus a lot of sensor data has to be processed at the edge, which raises security and privacy concerns. Additionally, there are different public and private organizations, who own the data, and do not want to share with each other the data in raw form for computations. This needs using Privacy Preserving Machine Learning (PPML) tools to analyze data across owners. The research work has been done by PhD students Ismi Abidi (2019-2023), Mehreen Jabben (2020-) and MTech Student Gauri Gupta (2022-2023). Ismi is currently Senior Lead Engineer in Qualcomm Hyderabad, and Gauri is pursuing PhD at MIT.
[ASIACCS24] "SpotOn: Adversarially Robust Keyword Spotting on Resource-Constrained IoT Platforms", Mehreen Jabbeen, Vireshwar Kumar, Rijurekha Sen in the 19th ACM ASIA Conference on Computer and Communications Security (Core A)
|
[PoPETS23] "End-to-end Privacy Preserving Training and Inference for Air Pollution Forecasting with Data from Rival Fleets", Gauri Gupta, Krithika Ramesh, Anwesh Bhattacharya, Divya Gupta, Rahul Sharma, Nishanth Chandran, Rijurekha Sen in Proceedings of The 23rd Privacy Enhancing Technologies Symposium (Core A) |
[NDSS22] "Privacy in Urban Sensing with Instrumented Fleets, Using Air Pollution Monitoring As A Usecase", Ismi Abidi, Ishan Nangia, Paarijaat Aditya, Rijurekha Sen in Proceedings of The Network and Distributed System Security Symposium (Core A*) |
[ACSAC21] "Practical Attestation for Edge Devices Running Compute Heavy Machine Learning Applications", Ismi Abidi, Vireshwar Kumar, Rijurekha Sen in Proceedings of the Annual Computer Security Applications Conference (Core A) |
Thread 3: Empirical measurement and design of system software for edge computing. Edge or IoT platforms have undergone an exponential growth in recent times, with Field Programmabale Gate Arrays (FPGA), GPU and custom ASIC chips integrated in these platforms to support AI-ML applications, as well as to enhance security. The system software stack is growing at the same rate, to bridge the gap between fast evolving applications from the AI-ML/security research communities and the newly integrated hardware features coming from the Computer Architecture research community. This third research thread was to understand this rapidly evolving eco-system, based on empirical measurements with recent edge platforms. Some design and implementation of system software was also done. The research was done by PhD students Shikha Goel (2018-2024) and Omais Shafi (2020-), and post-doc Mohammad Khalid Pandit. Shikha is currently in NVIDIA Bengaluru GPU performance team, Omais is in Qualcomm Bengaluru, and Khalid is Assistant Professor in NIT Hamirpur.
[ACMSEC23] "Bang for the Buck: Evaluating the cost-effectiveness of Heterogeneous Edge Platforms for Neural Network Workloads", Amarjeet Saini, Omkar B. Shende, Mohammad Khalid Pandit, Rijurekha Sen, Gayathri Ananthanarayanan in the 8th ACM/IEEE Symposium on Edge Computing (Core unranked) |
[JETC23] "Repercussions of Using DNN Compilers on Edge GPUs for Real Time and Safety Critical Systems: A Quantitative Audit", Omais Shafi, Mohammad Khalid Pandit, Amarjeet Saini, Gayathri Ananthanarayanan, Rijurekha Sen in ACM Journal on Emerging Technologies in Computing Systems, (significantly extended version of IISWC21 paper). (ACM Journal) |
[FPT22] "EXPRESS: CNN Execution Time Prediction for DPU Design Space Exploration", Shikha Goel, Rajesh Kedia, M. Balakrishnan, Rijurekha Sen in Proceedings of the International Conference on Field-Programmable Technology (EE Conference, Core unranked) |
[COMPASS22_2] "DynCNN: Application Dynamism and Ambient Temperature Aware Neural Network Scheduler in Edge Devices for Traffic Control", Omais Shafi, Sachin Kumar Chauhan, Gayathri Ananthanarayanan, Rijurekha Sen in Proceedings of ACM SIGCAS and SIGCHI Computing and Sustainable Societies (Core unranked) |
[IISWC21] "Demystifying TensorRT: Characterizing Neural Network Inference Engine on NVIDIA Edge Devices", Omais Shafi, Chinmay Rai, Gayathri Ananthanarayanan, Rijurekha Sen in Proceedings of IEEE International Symposium on Workload Characterization. (Core unranked) |
[FPL21] "EnergyNN: Energy Estimation for Neural Network Inference Tasks on DPU", Shikha Goel, M. Balakrishnan, Rijurekha Sen in Proceedings of the International Conference on Field-Programmable Logic and Applications (EE conference, Core unranked) |
[FPT20] "INFER: Interference Aware Estimation of Runtime for Concurrent CNN Execution on DPUs", Shikha Goel, Rajesh Kedia, M. Balakrishnan, Rijurekha Sen in Proceedings of the International Conference on Field-Programmable Technology (EE conference, Core unranked) |