An anomaly detection model using an autoencoder and a random forest to detect and remove anomalous network traffic. The IoT packets were collected using a Raspberry Pi, and the model is also designed to run on lightweight devices like Raspberry Pis.
A comprehensive analysis of anomaly detection from sensor frequency in a smart environment using Isolation Forest. Tested and validated on 5 different datasets.
An optimized transformer-based image restoration method. This method integrates existing image restoration methods like SwinIR and ESRT to develop an efficient approach.
Integration of the PyMoo library with the counterfactual generation library DiCE ML. This project extends the existing DiCE library by integrating PyMoo functions within its main counterfactual generation class.
Fine-tuning a Llama-3-8b model on a vehicular communication log dataset to identify any vehicular misbehavior using natural language.
Classifying movie reviews as good, bad, or neutral using Naive Bayes. The preprocessing and Naive Bayes implementation were done from scratch.
To validate and verify the sensor grouping algorithms that use spectral clustering, we proposed 10 metamorphic relations and tested them on multiple datasets.
A simple retro racing game built using a Microcontroller, Gyro sensors, and a dot matrix.