Title:	Detecting Sybil and Fake Users in mobile crowdsensing systems
DNr:	Berzelius-2025-70
Project Type:	LiU Berzelius
Principal Investigator:	Cihan Eryonucu <eryonucu@kth.se>
Affiliation:	Kungliga Tekniska högskolan
Duration:	2025-02-24 – 2025-09-01
Classification:	10211
Homepage:	https://nss.proj.kth.se/
Keywords:

Abstract

Mobile Crowdsensing Systems (MCS) (e.g., Google Maps) rely on large-scale user participation for data collection, making them vulnerable to Sybil attacks where adversaries register multiple fake identities to manipulate system outcomes. This project aims to developt a machine learning-based framework designed to detect Sybil attacks in MCS. Our proposed system leverages combination of Dynamic Time Warping (DTW), Density-Based Spatial Clustering (DBSCAN), Variational Autoencoders (VAEs) and Recurrent Neural Networks with Long Short-Term Memory (RNN-LSTM) to identify adversarial registrations while preserving user privacy. The requested computing resources will be used for training and optimizing VAEs and RNN-LSTM models on large-scale datasets. These experiments will enable further refinement of the system to enhance security in open-participation MCS environments.