Priyanshu Yadav
About
B.Tech student in Computer Science and Engineering at DTU (2023–present) conducting research on backdoor attacks and detection in data-distillation pipelines. He studies poisoned-sample detection through feature-space analysis, layer-wise representation consistency, Mahalanobis distance and feature-similarity measures. His project work includes face anti-spoofing and a QuickDraw web game integrating deep-learning models.
Research focus · Backdoor Security in Data Distillation
Achievements
- Volunteer teacher with Help Sewa Foundation for six months.
Key Projects
Backdoor Attacks and Detection in Data Distillation (ongoing)
investigates vulnerabilities in distilled datasets and develops poisoned-sample detection methods using feature-space analysis, layer-wise representation consistency, Mahalanobis distance and feature similarity.
Face Anti-Spoofing Detection System
developed a MobileNetV2-based presentation-attack detection pipeline on the LCC FASD dataset, with augmentation and preprocessing for varied spoofing conditions.
QuickDraw Online Game
built a MERN application that uses CNN, MobileNetV2 and ResNet models to classify sketches from the Google QuickDraw dataset.
Language-learning speaking-practice project (announced)
co-announced a collaborative project focused on accessible conversational practice.