Debanshu Das

Debanshu Das

Innovating research on critical infrastructure security, one project at a time. Passionate about enhancing cybersecurity through advanced methodologies in process mining, attack graphs, and AI-driven threat analysis.

Sitemap

A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.

Pages

Posts

PALM: A framework to identify novel attacks in an e-commerce system

less than 1 minute read

Published:

The widespread adoption of e-commerce and its lucrative, decentralized, multi-agent nature have made these systems vulnerable to cyber-attacks. Traditional signature-based approaches have been successful in detecting instances of fraud, however continues to struggle against unknown attacks. To this end, we present PALM, a framework that effectively integrate contextual knowledge on known vulnerabilities and threats with data-oriented process-aware approaches to build a comprehensive graph-structure-based model. Graph Convolution Network is then employed to look for malicious feature correlations and information flows, indicating a likelihood of link between un-connected neighboring nodes. Using a sample event log, we test the efficacy of our framework, achieving 83.33% accuracy and an AUC-ROC of 0.9186, signifying its potential in identifying novel attack paths GitRepo.

academic_projects

AgeWise.ai

Published:

Developed AgeWise, a facial recognition project that predicts a person’s age range based on their facial features. Achieved competitive accuracy by applying superior Deep Learning approaches and the possibility to customize skincare products. The model was developed using Machine Learning and Deep Learning and hence we were able to detect dark spots, puffy eyes and wrinkles on images in order to understand ageing signs.

VLANfinity

Published:

Efficiently revamped our campus Networking structure and included configuration of VLANs on all switches, the assignment of network devices to specific VLANs, and the implementation of proper VLAN tagging was done.The primary goal was to create a network that is both secure and well-organized, while also maintaining high performance and availability. Users of all regions will have access to the server and can communicate without creating congestion in the Network.

EndemiCast

Published:

Epidemic models based on ordinary differential equations, which effectively describe dynamic systems in many fields of science. As part of this project, US population data was utilized to model the potential spread of the coronavirus. Integrated state-of-the-art machine learning techniques and state estimation algorithms to better understand the dynamics of the pandemic system.

CredSense

Published:

During the online application process, the company seeks to automate (in real time) the loan qualification process based on the information entered by the customer. ML models will help the company predict loan approvals, thereby speeding up the decision-making process for determining whether or not an applicant is eligible for a loan.

HR Planes

Published:

Designed a novel object detection model to automatically detect airplanes in high-resolution satellite images, using Google Earth ismagery under various landscape, seasonal, and satellite geometry conditions. The dataset was evaluated using two state-of-the-art object detection methods. Incorporated advanced techniques such as hyperparameter tuning, optimization algorithm, and data augmentation to improve model precision. Attained excellent accuracy of 80.27% with training. Intended to accomplish reliable and accurate airplane detection capabilities.

TransLingua

Published:

This enthralling project involves constructing a neural machine translation model with large-scale parallel corpora by using an intuitive method, sequence-to-sequence learning approach consisting encoder-decoder LSTMs(Long Short-Term Memory) network architecture, RNN(Recurrent Neural Network) and word embedding. The model secured an outstanding translation accuracy along with minimizing the loss function eminently.

DeepRumor

Published:

I researched an end-to-end framework for early rumor detection on social media using deep learning techniques. By leveraging a range of algorithms, including weak supervision, I was able to address key bottlenecks in previous rumor detection systems. My work resulted in significant improvements in state-of-the-art performance for early rumor detection.

publications

Simulating cyber-attack scenarios by discovering Petri Nets from large scale event logs

Published in 16th International Conference on Communication Systems &NetworkS (COMSNETS), 2024

Developed and implemented an innovative approach to cyber-risk evaluation, leveraging log files and Petri-Nets, bridging formal models with cybersecurity.

Recommended citation: M. D. Makwana, V. Thakkar, D. Das and R. Kumar, "Simulating Cyber-Attack Scenarios by Discovering Petri-Nets from Large-Scale Event Logs," 2024 16th International Conference on COMmunication Systems & NETworkS (COMSNETS), Bengaluru, India, 2024, pp. 49-54, doi: 10.1109/COMSNETS59351.2024.10427052.
Download Paper

PALM: A framework to identify novel attacks in an e-commerce system

Published in 29th IEEE Pacific Rim International Symposium on Dependable Computing (PRDC 2024), 2024

Implemented an innovative approach to evaluate cyber risks through a behavioural model of an e-commerce system and conducted the prediction of novel exploits potentially extending from the system.

Recommended citation: Kumar, R., Pandey, S. and Das, D., PALM: A framework to identify novel attacks in an e-commerce system..
Download Paper | Download Slides

talks

teaching

Teaching experience 2

Workshop, University 1, Department, 2015

This is a description of a teaching experience. You can use markdown like any other post.