Name: Mirambika Patel
Roll No. : 2410110201
Email ID: mp930@[Link]
Course: Computer Science and Engineering
Title: Finding the source of an outbreak in a dynamic social network
Introduction
The study of information propagation and influence dynamics in social networks has gained
significant attention in recent years, particularly in the context of understanding how rumours,
misinformation, and other forms of content spread through online platforms. This field
encompasses critical challenges in identifying the origins of information outbreaks, which is
essential for effective content moderation and maintaining the integrity of online communities.
Source identification plays a critical role in various domains, including patient zero tracing,
social network rumor source detection, and more, as it enables tracing propagation processes
and implementing effective measures to block transmission [6]. Unlike traditional static
network analysis approaches, modern social media environments are characterized by highly
dynamic networks that continuously evolve through changing user relationships, varying
engagement patterns, and temporal activity cycles. The complexity of these systems is further
amplified by the multi-modal nature of available data, including content semantics, user
behavioural patterns, and temporal engagements.
Extensive research has been conducted across various aspects of source identification and
information diffusion in social networks. Researchers have focused on finding information
diffusion's seed nodes using specialized degree centrality measures [5], while studies on
detecting rumor outbreaks in online social networks [4] have highlighted the challenges posed
by the rapid and accessible nature of social media communication. Li et al. (2023) present a
comprehensive survey on Influence Maximization, highlighting how Machine Learning
approaches, especially Deep Reinforcement Learning, offer improved solutions despite the NP-
hard nature of the problem. Liu et al. (2022) model rumour transmission using a susceptible,
hesitating, infected, mitigated, and recovered (SHIMR) framework, analyzing user sharing
behaviour through strategic game theory. Xiang et al. (2025) introduce an innovative
hyperparametric influence minimization model that controls diffusion through node feature
interventions, demonstrating superior performance through greedy-based algorithms that
exploit the problem's super modularity properties.
Motivation
Despite significant contributions to social network analysis and influence dynamics, critical
gaps remain in current research that limit the effectiveness of outbreak source identification.
Current approaches suffer from several methodological limitations including graph-centrality
methods that assume sources occupy central positions, statistical approaches requiring strong
distributional assumptions, and machine learning methods lacking interpretability across
different content types. Most critically, existing approaches fail to effectively leverage temporal
information and multi-modal data available in social media contexts, neglecting rich signals
that could significantly improve accuracy. There exists a notable absence of comprehensive
evaluation frameworks that assess performance under realistic conditions including incomplete
network observability, noisy propagation data, multiple simultaneous sources, and scalability
constraints for large-scale networks.
Objective
This research aims to develop a comprehensive framework focusing on two primary objectives:
multi-modal analysis and source outbreak identification in dynamic social networks. The multi-
modal analysis component will integrate diverse data sources including structural network
features, content semantics, user behavioral patterns, and temporal engagement signals to
create robust detection capabilities that perform effectively even with incomplete or noisy data.
The source outbreak identification component will develop dynamic algorithms that trace back
through time-evolving network structures to identify likely source nodes, accounting for
changing topology and temporal influence patterns while leveraging comprehensive temporal
information and network evolution patterns. The research will establish evaluation frameworks
for assessing performance under realistic conditions including multiple potential sources and
incomplete network observability, followed by real-world validation experiments to
demonstrate framework effectiveness and establish practical implementation guidelines for
social media platforms and content moderation systems.
Timeline
Time Task to be completed
1st Quarter Literature Survey
2nd Quarter Problem statement formulation
3rd Quarter Implementation of multiple models for
finding the source of the outbreak in a
dynamic social network
4th Quarter Research paper writing
(likelihood)
P(A|B) = P(B|A) x P(A) / P(B)
(posterior)
References
1. Li, Yandi, Haobo Gao, Yunxuan Gao, Jianxiong Guo, and Weili Wu. "A survey on
influence maximization: From an ml-based combinatorial optimization." ACM
Transactions on Knowledge Discovery from Data 17, no. 9 (2023): 1-50.
2. Liu, Wenjia, Jian Wang, and Yanfeng Ouyang. "Rumor transmission in online social
networks under Nash equilibrium of a psychological decision game." Networks and
spatial economics 22, no. 4 (2022): 831-854.
3. Xiang, Bin, Bogdan Cautis, Xiaokui Xiao, and Laks VS Lakshmanan.
"Hyperparametric Influence Minimization: Feature-Driven Intervention Beyond
Blocking." In Proceedings of the 31st ACM SIGKDD Conference on Knowledge
Discovery and Data Mining V. 2, pp. 3297-3308. 2025.
4. Kumar, Saurabh, et al. "Detecting rumor outbreaks in online social networks."
Social Network Analysis and Mining 13, no. 1 (2023): 67.
5. Zhang, Wei, et al. "Finding Information Diffusion's Seed Nodes in Online Social
Networks Using a Special Degree Centrality." SN Computer Science 5, no. 4
(2024): 1-15.
6. Peng, Shaoning, et al. "Detection of Rumors and Their Sources in Social Networks:
A Comprehensive Survey." arXiv preprint arXiv:2501.05292 (2025).