Oracle: Working as part of the observability team and developed solutions to provide better observability for OCI clusters and microservices.

Tools and Technologies used: Golang, Kubernetes, Microservices, Terraform, OpenTelemetry, Prometheus, Grafana, Loki, Mimir, Alloy, Tempo.

CSE - University of Nevada Reno: Worked as a tenure-track Assistant Professor, directed the IoTSec Laboratory and supervised 4 PhD students. I primarily led two projects (detailed below), one of which received an NSF award.

• Network-Aware Edge Computing for Real-time Wildfire Detection — Developed an edge computing framework tailored for delay-sensitive scientific applications, such as wildfire monitoring. This framework uses tools such as iostat, dstat, mpstat, monit, and sar to monitor resources and perform application profiling such as CPU and memory utilization. Then using these collected data we implemented Mixed Integer Linear programming, and Genetic algorithm based resource allocation. Project focused on two specific applications that can benefit our AlertWildfire framework. First we explored latency benefits when we perform object detection at edge rather than cloud. Second we explored bandwidth benefits when we downscale image data at edge prior to sending it to cloud and upscaling in cloud for storage. This project received $500K NSF funding (link to award). (Lead the effort)
• Self-Learning and Lightweight Network Traffic Fingerprinting for IoT networks — In this project, we focused on network traffic fingerprinting, a technique used to identify devices in a network and perform intrusion detection. Traffic fingerprinting analyzes network traffic characteristics of devices (we primarily focused on IoT devices) and extracts traffic features (e.g., packet size, inter-arrival timing) and feed these features into machine learning models for classification and anomaly detection. We addressed two main challenges. First, machine learning-based traffic fingerprinting can suffer from concept drift. To mitigate this, we developed self-learning machine learning models that can automatically adapt and re-tune hyperparameters when concept drift occurs. Second, traffic fingerprinting models need to be deployed in resource-constrained environments such as routers and edge servers. Since these models rely on computationally expensive machine learning techniques, optimizing their overhead is crucial. To achieve this, we developed optimization methods based on genetic algorithms, grey wolf optimization, and bee colony optimization to fine-tune the underlying machine learning models by balancing accuracy and computational overhead. (Lead the effort)
  Articles: A

Tools and Technologies used: Python, scikit-learn, tensorflow, tensorflow-light, IoT, SDN, network telemetry, NVIDIA Jetson edge servers, wireshark, tshark.

Systems Engineering Research Center: Worked as a networking and machine learning researcher. Following are some of the projects that I was involved.

• IoT device identification system — utilizes machine learning with network traffic of the devices to identify which IoT devices attached to the network. We also developed locality sensitive hashing based system which will enable to bypass complex feature selection, hyper parameter tuning and retraining the machine learning based methods. (Lead the effort)
  Articles: A, B, C, D
• Intrusion detection systems for IoT devices — system monitors traffic flow of the devices and uses one-class classifier and autoencoder models to detect malicious activities. (Lead the effort)
  Articles: A, B, C, D

Tools and Technologies used: Python, scikit-learn, tensorflow, wireshark, tshark, numpy, pandas, matplotlib.

UNR Cybersecurity Center: Worked as a networking and edge computing researcher. Following are some of the projects that I was involved.

• RIVA — is data transfer system for high performance networks. It is main benefit compared to state of the art Globus is providing integrity of transferred data against silent data corruptions (these corruptions happen when we write data from memory to disk.) (Lead the effort)
  Articles: A, B
• Complex networks of systems — applied network science to analyze complex systems, such as US migrations and air-transportation systems. (Lead the effort)
  Articles: A, B

Tools and Technologies used: C, C++, Java, TPM, Gephi, Pajek.