Coded Computation for Internet of Things

The Internet of Things (IoT) is emerging as a new Internet paradigm connecting an exponentially increasing number of smart IoT devices and sensors. IoT applications include smart cities, transportation systems, mobile healthcare and smart grid, to name a few. Unlocking the full power of IoT requires analyzing and processing this data through computationally intensive algorithms at unprecedented high rates, with stringent reliability, security and latency constraints. In many scenarios, these algorithms cannot be run locally on the computationally-limited IoT-devices and are rather outsourced to the cloud. This leaves the IoT network, and the applications it is supporting, at the complete mercy of an adversary, or a natural disaster that can jeopardize the IoT, or completely disconnect it from its “brain” (the cloud), with potentially catastrophic consequences.

As a solution to mitigate the computational bottleneck in IoT, we focus on the scenario that IoT-devices help each other in their computations in a distributed fashion, with possible help from the cloud, if available. Our approach is based on the new theory of coded computations, which studies the design of erasure and error-correcting codes to improve the performance of distributed algorithms through “smart” data redundancy. We have developed distributed, adaptive, and secure coded computation algorithms for IoT devices and deployed these algorithms on real IoT devices to demonstrate the efficiency of the approach. This work has been supported by Army Research Lab (ARL) #W911NF-1820181 and #W911NF-1710032, National Science Foundation (NSF) #CNS-1801708, Army Research Office (ARO) #W911NF1810211 grants.

Modeling and Development of Resilient Communication

ReDiCom 

Effective communication among first responders during and in the aftermath of a disaster can affect outcomes dramatically. We seek to build a resilient architecture that allows first responders to communicate even with: (i) damage to infrastructure – civilian and/or specialized communication facilities may be damaged by the disaster, (ii) congested channels – because affected people report something about the disaster, and these messages may be broadcast, (iii) dynamically formed groups – first responder teams may be formed dynamically in response to a disaster and team member addresses (e.g., phone numbers) may not be known to one another, (iv) impediments to communication – because the new command chain to manage the disaster may be different from the original organizational hierarchy,(v) poor interoperability – each sub-team might use different communication facilities, and (vi) security attacks – disaster situations are often vulnerable to attacks, requiring authentication and authorization as well as establishing data integrity and provenance. This work is supported by NIST grant #70NANB17H188. See our project webpage.

Transport Protocols for Internet of Things

IoT Transport 

As Internet of Things (IoT) adoption grows, it will have a significant impact on cellular networks, both access and core. Potentially billions of devices with vastly different throughput, latency, and signaling requirements will communicate over cellular networks. Though typically thought off as low-traffic volume devices, IoT devices have interesting characteristics that can cause them to result in increased load on the network. The goal of the project is to develop transport layer and scheduling mechanisms that differentiates IoT traffic from the others and make prioritization accordingly. This is a collaborative work with Dr. Vamanan from the CS department of UIC, and Drs. Gopalakrishnan and Halepovic from AT&T Labs Research.

Anomaly Detection in Video Using Computationally Efficient Machine Learning Algorithms

Anomaly Detection 

Anomaly detection is crucial for surveillance applications. Although humans can easily interpret normal interactions in a video and determine if an activity is unusual or not, this is a challenging task for machines. The goal of this proposal is to develop computationally efficient machine learning algorithms for detecting unusual events in surveillance videos. In particular, we design, develop, and use computationally efficient generative adversarial networks (GAN). This is a collaborative work with Prof. Cetin and Prof. Koyuncu from the ECE department of UIC, and supported by two Seagate grants.