The CSE department faculty is actively involved in research in various fields of Computer Science. The department provides an excellent research platform, and nurtures and challenges students to solve real-world research problems. The research in the department can be broadly classified into the following areas.
We are a group of faculty and students working on exciting problems on the recently very popular areas in algorithms and data structures including dynamic graph algorithms, fault tolerance, streaming algorithms, computational geometry, graph theory and space bounded algorithms. The groups has been carrying out excellent research. In the last 5 years (2014-2018), the group has published 5 ICALP papers including one best student paper in 2016, 1 paper in STOC, 3 papers in SODA. ACM Distinguished Dissertation Award for the year 2019 was also awarded to a PhD thesis in algorithms at IIT Kanpur. We are interested primarily in the research problems and general directions mentioned above, but are also adaptable and receptive to new interesting problems that may come up in the near future.
Our research lies at the unique intersection of data science, visualization, machine learning, high-performance computing, computer graphics, and human-computer interaction. We develop cutting-edge data analytics, visualization, and machine/deep learning-driven solutions to accelerate discoveries in diverse application domains. Summarizing, extracting, and comprehending the crux from the vast seas of data and representing them visually and interactively in an interpretable manner is the broad-scale focus of our research.
We study how machine (deep) learning techniques can be used to solve challenging big data visual computing problems, termed ML4VIS. In contrast, we also employ interactive visual analytics techniques to build efficient tools that can be used to comprehend how the complex BlackBox machine (deep) learning models work to further research in the areas of interpretable and explainable machine learning via visual analytics, termed VIS4ML. To perform scalable visual analysis of extreme-scale multifaceted scientific data, we use modern HPC capabilities to develop high-performance data analysis methods and conduct in situ analysis and visualization research for their wide applicability in diverse scientific domains.
We report our research results in various high-quality journal and conferences, e.g., IEEE TVCG, Computer Graphics Forum (CGF), IEEE Visualization, EuroVis and EuroGraphics, ACM CHI, IEEE Pacific Visualization, IEEE BigData, IEEE LDAV, EGPGV, etc.
Modern biology is in the middle of a paradigm shift where computation is an essential taskforce for understanding biological systems. Computational biology is a broad discipline that aims to develop novel computational methods utilizing elements from a wide range of mathematical and computational fields (e.g., algorithmics, machine learning, statistical inference, etc.) for building models for diverse types of experimental data (e.g., sequences, images) and understanding biological systems (e.g., cells, tissues, organs, etc.). The Computational Biology group at IIT Kanpur focuses on the design of scalable computational techniques backed by probabilistic modeling and statistical inference methods for understanding the biology of cancer and processes in single cells. Using their computational frameworks, the group tries to understand how cancer cells evolve, elucidate the heterogeneity in cancer tissue, and identify potential drug targets. For this, the group also collaborates with faculties in the BSBE department as well as different international labs. For single-cell sequencing, the group develops machine learning algorithms for problems such as cell lineage reconstruction, cell type clustering, trajectory inference, gene regulatory network inference, etc.
The group actively collaborates with international academic labs (e.g., in Rice University, UT MD Anderson Cancer Center, etc.) as well as industry (e.g. Mission Bio) to work on cutting edge problems. Previous research works from the group have been published in top research journals such as Nature Methods, Genome Biology, etc.
Data drives the industry 4.0. The importance of data in our current technological landscape is aptly embodied in the phrase, “Data is the new oil.” As more and more different facets of our society such as financial, social, educational, etc., are becoming digital, we are producing data at an unprecedented scale. This massive amount of data is consumed and processed at multiple levels by multiple organizations all around the world to hyperdrive innovations in digital technology to new heights.
However, this data-driven world is also filled with gross malpractices. It is not uncommon to misuse private and sensitive data, often from unsuspecting and gullible users, for guerilla marketing, mass manipulation, oppressing vulnerable sections of society, influencing democratic processes, etc. Therefore, we need regulations, standards, and techniques for secure and ethical data handling. Cryptographic techniques such as homomorphic encryption, functional encryption, multi-party computation, often denoted by the umbrella term computation on encrypted data (COED), and privacy engineering techniques such as zero-knowledge proofs facilitate ethical data processing while preserving its privacy. Such methods give back the control of the data to its original owners.
We actively research computation on encrypted data and privacy engineering, showcasing our commitment to advancing information security. Exploring innovative cryptographic techniques, such as homomorphic encryption, we aim to enable secure data processing while safeguarding confidentiality. Our research contributes to developing efficient computations and protecting sensitive information. We believe in practical demonstrations. If data is the raw material of Industry 4.0 then the computation power is the machinery that processes this raw material. So, we must be prudent while extracting maximum efficiency from our hardware to make these schemes suitable for practical use. In summary, our work spans both theoretical and practical aspects of COED and privacy engineering techniques. We publish our works regularly in highly reputed and competitive conferences and journals. We also have active collaboration with multiple leading research groups and industries worldwide working in these domains.
The Computer Architecture and Operating Systems group focuses on all aspects of modern computing systems: processor, memory, storage, and interaction with the operating systems. Some of the active topics of interests are the following: better processor design with speculative techniques, memory hierarchy optimizations in the form of hardware prefetching, cache/DRAM content management, and cache coherence protocols for client and server systems. The group works on security issues related to modern processor, memory systems, and operating systems, in the form of side/covert channel attacks and their mitigations. The group also focuses on the interface between Operating Systems, Architecture, Virtualization, and Cloud Computing.
The group is consistently publishing in some of the top conferences and journals like ISCA, MICRO, HPCA, PACT, ICS, VEE, MIDDLEWARE, and TACO. The group currently is funded by Intel, Nutanix, semiconductor research consortium (SRC), and UP defense corridor projects.
Cryptography is the science of protecting our information in an untrusted environment. Due to rapid digitization of many facets of our life, it has become utmost important to protect those data from people or organization with evil intent.
In our group, we study cryptographic or computer security techniques from both theoretical and practical perspectives. So, you may be passionate about writing fast, secure, and extremely frugal or massively parallel code, or find pleasure in designing and breaking secure protocols, or love to see the esoteric mathematical concepts you learned to take shape in something to solve problems in our real world, we have projects and problems to excite all of you. We constantly engage in designing, analysing and breaking cryptographic primitives such as digital signatures, secure key establishment protocols, etc using mathemetical analysis or practical attacks such as fault, side-channel or micro-architectural attacks. We also implement these protocols on various platforms such as microcontrollers, GPUs, FPGA, ASICs, etc. We have vast collaboration national and international collaboration where we work jointly on projects, arrange student internships and short visits.
In our current digital ecosystem, expertise in computer security and cryptography are extremely valuable and rare skill. This problem is even exacerbated by the fact that the landscape of this field is constantly changing. So one needs to continuously update himself or herself with the latest technologies in this field. In our group, we always try to chase excellence and strive to work on the state-of-the-art techniques and tools to stay relevant. Our current focus is on post-quantum cryptography, homomorphic and functional encryption, homomorphic friendly ciphers, 5g/6g ciphers, microarchitectural, side-channel attacks and fault attacks. As a testimony of our work, we regularly publish our works on highly reputed and competitive conference and venues. We also have strong collaboration and strong exchange programs with many national and international industry and research groups like COSIC at KU Leuven, Belgium, Arm Research at Cambridge, UK, NTT coroporation at Japan, SEAL lab at IIT Kharagpur, SPARC lab at Purdue University, USA, etc.
In this age of automation, usage of complex safety-critical systems are increasingly high in industries like hardware, automobiles, avionics, space etc. This is a major task to ensure correctness and completeness to get a reliable controller of such systems. The CPS Group at the CSE Department of IIT Kanpur is working on developing principled approaches for robust implementation of cyber-physical systems. A cyber-physical system is a collection of interconnected computing devices interacting with the physical world to regulate its behavior. The group is working on several exciting problems in the area of CPS, more specifically in distributed multi-robot systems and internet-of-things.
The focus of this group is to develop cyber-physical systems with correctness guarantee through the application of formal verification/synthesis techniques.
Today, cyber-Security is not just a computer science issue but a national security issue. Terrorist organizations like ISIS use digital platforms to recruit fighters while many countries including China, Russia, Israel and the United States have cyber-attacked rival nations. For example, Operation Aurora by China stole intellectual property of a number of American companies. The American/Israeli Stuxnet worm crippled Iran’s nuclear program. Like many other nations, India’s national critical infrastructures are ripe targets for cyber-attacks. At IIT Kanpur, leveraging the expertise in multiple areas of Computer Science and Engineering, we initiated such a comprehensive program within our center for cyber-security. We have a multi-disciplinary national project to carry out research, training and education in cyber security of the national cyber space including information infrastructure, and other critical infrastructures such as banking, power grid, industrial manufacturing, defense tactical communication networks, and various information assets of the country.
Our approach is multi-pronged and multi-layer -- defense-in-depth strategy. Starting from crypto algorithms, protocols, till the systems. We also use machine learning for anomaly detection to fight persistent threats in the critical infrastructure.
Several cyber security research groups from IIT Bombay, IIT Kharagpur, ISI Kolkata, IIIT-Delhi, and MNIT, Jaipur have agreed to associate their research groups/centers with the proposed center. Exchanging faculty and study for short and long periods for collaborative research and pursuing problem specific funding together will be the modality of these associations. As a showcase of the center's productivity, three specific problem areas have been identified -- solutions of which are being pursued and show cased to the highest levels of the government.
Databases are ubiquitous and form the backbones of almost all the modern systems. While traditional databases are relational, the NoSQL paradigm has proved itself useful in various applications and situations, especially in the realm of big data. The research in this field encompasses data analytics, data processing, indexing, querying, searching and information retrieval. It also includes data mining in the form of graphs, text, multimedia, strings, etc.
With proliferation of data emerging from heterogeneous sources, data analytics is being considered as the key tool for business growth, and understanding the physical world. While there are already many tools for analyzing structured data, there are still many open problems for unstructured or semi-structured data, speech, and video data. The “Big Data” phenomena encompasses all of these, and is thus the buzz word of this era.
Over the past few decades, Internet use has grown precipitously worldwide. Today, over 4.7 billion people use social media; we go online for music, news, television, and movies and communicate with family and friends; essential day-to-day services like shopping, banking, and even health care are increasingly delivered virtually. Further, the COVID-19 pandemic has only accelerated these trends. On the one hand, shifting to online services increases efficiency and convenience; on the other hand, it has created an ecosystem of surveillance capitalism riddled with severe privacy threats. One way of protecting data is building systems such that no unauthorized entity can access the data they are not supposed to. For instance, building firewalls is one option. A more robust approach to tackling this problem is making systems that are data-leak proof, i.e., even if an authorized entity gets access to the data they are not supposed to, they cannot make any sense of it. Our research is in creating systems for the secure and private processing of data, ensuring that data leaks cannot happen, even if an adversary penetrates a company’s network. In our research, we build data-leak-proof systems. In creating such data-leak-proof systems, we develop cryptographic primitives and tools that have applications beyond the particular system. There are three broad approaches to building data-leak-proof systems:
The three aforementioned methods of achieving secure and private data processing have different pros and cons. For instance, while homomorphic encryption performs the worst, trust is the most reliable. Putting one’s trust in secure hardware is less reliable, as some recent works have shown, but it leads to good performance. Distributed trust is a middle ground in settings where a trust assumption is reasonable. In such settings, the protocols are secure so long as the parties involved obey the non-collusion assumption. These lead to what has been described as “probably private” protocols.
In its broadest strokes, our research aims to develop tools that help users protect their privacy online without sacrificing the modern Internet’s conveniences. Our research contributions run the gamut from low-level cryptographic innovations through the design and analysis of complex systems all the way to heavily optimized implementations of primitives and systems alike.
To make progress in this area some of the privacy-preserving technologies that we have worked on are a) Private Information Retrieval, b) Secure Multi-party Computation, c) Zero Knowledge Proofs, and d) Oblivious Random Access Memory.
Formal methods are an area of computer science where we use automated mathematical reasoning techniques to analyze programs. This analysis could be used to prove a program has certain correctness properties -- for example, that it does deadlock, or that it does not have a buffer overflow vulnerability, or to automatically find violations of these correctness properties. Our group applies formal methods in two distinct but overlapping areas: (i) for analyzing cyber-physical systems and (ii) proving security of systems.
Our formal methods groups combines expertise in robotics, cyber security and formal methods to solve real-world verification problems.
The department has very recently started exploring the area of hardware security that encompasses requirements of secure hardware design for novel cryptographic algorithms in FPGA/ ASIC, hardware design principles for real-time, low power embedded systems, side-channel (power, EM, and timing) and fault attacks and countermeasures, hardware fingerprinting with physically unclonable functions, hardware Trojan Horse design and detections, hardware IP Protection, development of quantum secure public key implementations, protocol design principles to bridge the gap between the hardware primitives and their secure use in applications, FPGA based hardware accelerator for AI and its security aspects such as privacy leakages, adversarial attacks etc. Currently, we are focusing on developing a testbed to integrate unconventional hardware security primitives with Micro-Air vehicles to realize a privacy-preserving anonymous authentication scheme for trusted data communication with the ground control system. We will also be investigating whether these hardware-assisted tools at the device level can be used as a countermeasure to security vulnerabilities at higher layers of the communication protocol stack. Development of low cost side channel evaluation platform along with an online side channel leakage detector is currently undergoing in the department. This setup would be useful for developing secure and efficient implementations of cryptographic algorithms.
Parallel programming is ubiquitous in today's multicore era and is a necessity to exploit performance from multicore architectures. However, massive parallelism entails significant hardware and software challenges. High performance computing is used to solve many real-world scientific problems. With the ever-growing compute capabilities, complex memory hierarchies, and varied network topologies, writing efficient and scalable parallel code is a significant challenge. The group focuses on several challenging problems in this area, such as topology-aware mapping, communication-aware job scheduling, and effective parallelization strategies. The group also looks into compiler optimizations for high-performance computing and domain-specific optimizations for applications like deep learning networks.
In today's era of big data, high performance computing can significantly speed up big data related computations and analysis. However, the challenge lies in big data I/O. The group also focuses on solving parallel I/O bottlenecks.
This is an area where the department has had a recent surge in terms of strength, as well as diversity. We cover nearly all the challenging areas of machine learning (ML) and computer vision. Research In ML focuses on probabilistic machine learning, deep learning, optimisation, statistical learning theory, natural language processing etc. Research in computer vision is mostly on language and facial analysis, graphics, human attributes prediction, pose estimation, and action/activity prediction. We also conduct research in human-centered computing, human factors in computing, and computational cognitive science to bring AI algorithms closer to humans both conceptually and instrumentally. An emerging area of research in AI is computational social choice that uses tools from CS and Economics to help humans take provably efficient decisions. Our department has faculty and students working in this interdisciplinary area.
Another interdisciplinary area of research in machine learning that is covered in the department is Computational Biology where the group develops novel probabilistic models and inference algorithms for understanding biological systems.
Results from this research group are frequently reported in ICML, NIPS, AAAI, IJCAI, AISTATS, CVPR, AAMAS, WINE, Games and Economic Behavior etc. The department has an active reading group in machine learning called SIGML which regularly hosts invited talks and guest lectures.
Developing machines that could seamlessly assist and augment human capabilities has been a holy grail for the field of artificial intelligence. Exploration Lab at the IIT Kanpur is working towards this goal. The group focuses on three main themes: Natural Language Processing (NLP) and Natural Language Understanding (NLU), Understanding and Modeling Human Behavior, and AI for Social Good.
Under the NLP and NLU vertical, the group works on various applications related to understanding and modeling human languages in various domains. For example, to streamline and make the Indian legal system more efficient, the group works on Legal-NLP: modeling legal documents for various applications, inter alia, judgment prediction and explanation, semantic segmentation of legal documents, prior case retrieval, etc. Other NLP applications include natural language interfaces to databases, bio-medical NLP, and ethical reasoning.
AI-based technologies are slowly becoming part of our lives; however, if AI-based machines do not understand human behavior, then they may not be able to communicate seamlessly with humans. In order to address this gap, Exploration lab has been working on developing algorithms for understanding human behavior and how it affects the decision-making process. In particular, the group works on Multimodal Multilingual Contextualised Affect Understanding models. The group also explores emotion and its causes and how it affects the decision-making process in humans. On the affect generation side, the group works on Affective Dialog Generation and has recently been exploring developing agents that could generate affective visual and textual responses based on the situation.
The group has also been exploring developing environments for imbibing common sense knowledge, and decision-making in the real world to deep reinforcement learning (RL) based agents. These environments are in the form of games based on real-life scenarios. Goals in these games are related to performing everyday activities like planting a tree, making coffee, repairing a bike, etc. RL-based agents play these games and learn from scratch about the world via trial and error. Similar to a typical RL setting, the agent receives observations from the environment and consequently performs actions that would take it closer to the goal and, in return, gets rewards from the environment. The research is aimed towards Embodied AI and for transferring knowledge to robots that could be deployed in the real world.
The last research theme that Exploration lab is exploring is solving societal problems using AI-based technologies. For example, the group works on sign language understanding and generation; another area is related to AI for mental health, where the group is studying the correlation between speech, language, and neuroimaging data for the diagnosis of Schizophrenia.
Exploration lab actively collaborates with the Industry and Academics both in India and outside India. The group has been actively publishing research at top-tier venues like ACL, EMNLP, COLING, EACL, TACL, AAMAS, ICMI, etc.
The department has a focused interest on the theory and applications of programming languages, program analysis and compilers research. The main research areas are compilers, data flow analysis, heap analysis, formal techniques for automated debugging, program verification and synthesis, software testing, static and dynamic program analysis, concurrency, high-performance compiler optimizations, and GPU algorithms. Of particular interest is a recent project being executed by Prof. Amey Karkare, Prof. Subhajit Roy and Dr. Sumit Gulwani (adjunct faculty, MSR Redmond) on developing intelligent tutoring systems [link] that are designed to adaptively guide students who are learning programming or other tools and tasks for the first time. The project has been successfully piloted with the introductory programming course at IIT Kanpur which graduates more than 800 students each year.
The group actively collaborates with researchers from other institutes and focuses on publishing in top PL and Systems venues OOPSLA, ICSE, ESEC/FSE, CC, EuroPar, ACM TACO, and ISMM among others.
Faculty: Dr. Amitangshu Pal
Our research interests span a seeming widely diverse set of topics, with the aim of developing adaptive solutions and techniques for building communication and sensing infrastructure for future IoT platforms. To be specific, our group's interests lie in three broad areas, i.e. (a) sensing, (b) communication and (c) networking. For (a) we are interested in building different IoT-based sensing platforms that can be useful for smart healthcare, agriculture, surveillance, transportation etc. For (b) we are exploring communication possibilities in different challenging environments, such as underground, underwater or inside body-area networks etc. These environments are challenging especially because the RF communication does not work in these media, and therefore exploring other possibilities like acoustic, magnetic or visual light based communication are promising. For (c) we explore networking solutions for wireless and optical networks, develop adaptive schemes for rechargeable sensor networks, extend networking solutions in challenging environments such as building disaster recovery networks, explore sensing and networking in underground pipeline environment, develop cyber-physical solutions in the fresh food logistics and smart city context etc. We also study several problems on content centric networks, reconfigurable data centre networks, enterprise networks etc.
Our group’s philosophy is to pick up important real-world problems, come up with accurate or tractable analytical models to represent them, then use relevant theoretical techniques to design optimal/near-optimal algorithms to solve these problems and finally, experimentally or by simulations, verify and analyse the proposed schemes. We work in multiple inter-disciplinary areas that span both algorithm design as well as system prototyping.
Our group has expertise in the use of formal methods to build provably secure systems. Recent research includes work on provably secure enclave platforms which can enable secure computing in the cloud and the provably secure defenses against transient execution attacks like Spectre and Meltdown. The group also focuses on security issues related to Computer Architecture, Operating Systems, and Computer Networks. Side-channel attacks at caches and processor, side-channel attack proofs, formal guarantees, crypto engineering to vulnerability analysis of systems and application layer software, network and web security, cloud security through virtual machine hardening.The group currently is funded by semiconductor research consortium (SRC), and UP defense corridor projects.
The department continues to be one of the best places to engage in cutting edge research in all areas of complexity theory, logic, game theory etc. Our graduate students and alumni have performed marvelously at arriving at path-breaking results on the very fundamentals of computer science. Recent achievements include a best student paper award at ICALP 2016 (the best paper award was also won by CSE IITK alumni), 3 papers at the premier venue STOC 2016, 3 acceptances at MFCS 2016, 2 at STOC 2018, and acceptances almost every year at CCC/ ICALP/ ISSAC/ STACS/ FSTTCS.
Thrust areas in the department include streaming algorithms (Prof. Sumit Ganguly), information theory (Prof. Satyadev Nandakumar), quantum algorithms and complexity (Prof. Rajat Mittal), game theory (Prof. Sunil Simon, Prof. Swaprava Nath), logic (Prof. Anil Seth), computational complexity theory (Prof. Manindra Agrawal, Prof. Nitin Saxena, Prof. Raghunath Tewari), and number theory/ algebraic algorithms/ algebraic complexity (Prof. Nitin Saxena).