From Perpetrators' Digital Footprints to Behavioral Communities: An Ontology-Driven Framework for Bengali Cyberbullying Analysis--- FULL VERSION

Md. Nesarul Hoque; Md Hanif Seddiqui; Miskatul Anwar; Rudra Pratap Deb Nath

cs.CLOct 22, 2026

From Perpetrators' Digital Footprints to Behavioral Communities: An Ontology-Driven Framework for Bengali Cyberbullying Analysis--- FULL VERSION

Authors: Md. Nesarul Hoque, Md Hanif Seddiqui, Miskatul Anwar, Rudra Pratap Deb Nath

Organizations: Big Data, Information and Knowledge Engineering Lab, Department of Computer Science and Engineering, University of Chittagong, Chattogram, 4331, Bangladesh · Department of Computer Science and Engineering, Gopalganj Science and Technology University, Gopalganj, 8105, Bangladesh

Abstract

Cyberbullying involves behavioral and contextual patterns beyond isolated abusive messages. However, existing approaches remain fragmented across content detection, semantic representation, and network analysis, with limited interpretable perpetrator representations, multi-scale structural modeling, and explanations of behavior-driven similarity and influence propagation. These limitations hinder the identification of persistent perpetrators, characterization of behavioral communities, and assessment of potentially influential users. This paper proposes Ontology-Driven Cyberbullying Behavior and Network Analysis (ODCBNA), a perpetrator-centric framework integrating requirements-driven ontology design, knowledge graph construction, behavioral modeling, community analysis, and influence modeling. Using the Bengali Contextual and Diverse Cyberbullying (BenCoDiC) corpus as a use case, ODCBNA identifies $1{,}063$ users with at least three bullying comments or replies as perpetrators and represents each through a $64$ -dimensional behavioral and contextual feature vector. Cosine similarity and $k$ -nearest-neighbor modeling produce a weighted network with $27{,}554$ edges. Louvain analysis identifies six communities with a modularity of approximately $0.495$ and $27$ sub-communities. Feature-contribution analysis identifies the Video, Impact, and User Profile groups as the strongest contributors to behavioral similarity and community separation, while contextual dimensions support interpretation. Community-constrained influence analysis shows that IM Greedy achieves small positive gains in all six communities, although different seed-selection methods produce comparable diffusion spreads despite limited seed overlap. These findings provide an interpretable basis for understanding perpetrator behavior, community organization, and potential influence patterns in Bengali online environments.

Figures & tables

Ref.	Semantic / Structural Basis		RG2	RG3		RG4
Ref.	Ontology	Perp-Graph	Behav. Rep.	Similarity	Community	Explainability	Influence
Ejaz et al. [ 13 ]	✗	✗	✗	✗	✗	✗	✗
Harshitha et al. [ 28 ]	✗	✗	✗	✗	✗	✗	✗
Cuzzocrea et al. [ 7 ]	✗	✗	✗	✗	✗	✗	✗
Xiong et al. [ 37 ]	✗	✗	✗	✗	✗	✓	✗
Atif et al. [ 2 ]	✗	✗	✓	✗	✗	✗	✗

Table 1: Comparative analysis of existing studies and ODCBNA across semantic, behavioral, structural, and analytical dimensions.

Figure 1: Overview of the ODCBNA framework.

Figure 2: High-level structure of the cyberbullying application ontology (TBox), including subclass relationships and selected datatype properties.

Figure 3: Ontology-driven knowledge graph construction pipeline.

Figure 4: RDF-to-heterogeneous-graph transformation process.

Figure 5: Perpetrator-centric subgraph induction from the heterogeneous knowledge graph.

Figure 6: Behavioral and contextual feature construction over the perpetrator-centric subgraph.

Network	#Nodes	#Edges	Avg Deg	Density	Clust	Path
Perpetrator	1063	27554	51.84	0.0488	0.1939	3.4192
Community	6	15	5.00	1.0000	1.0000	0.0015
Sub-community	27	311	23.04	0.8860	0.8932	0.0156

Table 2: Structural properties of the perpetrator, community, and sub-community networks.

Comm.	#Nodes	#Edges	Density	Avg Deg	Clust	Meta Deg
C0	213	3214	0.1424	30.18	0.2676	5
C1	141	2135	0.2163	30.28	0.2984	5
C2	184	3728	0.2214	40.52	0.3208	5
C3	104	1514	0.2827	29.12	0.3621	5
C4	237	4010	0.1434	33.84	0.2570	5
C5	184	3556	0.2112	38.65	0.3157	5

Table 3: Community-level structural properties of the perpetrator network.

Parent	#Subcomms	Avg Size	Largest Size
C0	4	53.25	64
C1	4	35.25	51
C2	6	30.67	42
C3	4	26.00	38
C4	5	47.40	63
C5	4	46.00	57

Table 4: Hierarchical sub-community structure within the six parent communities.

Comm.

Perp.

Act.

Act./P

Bully

B/P

Language

Target

Party

Time

Channel

User Profile

Video

Impact

213

1099

5.16

864

4.06

Bengali

Individual /

AL /

Afternoon /

Rtv News /

High Video /

High Avg. Comment /

High % Reply /

General People

Other

Morning

Ekattor TV

High View

High Avg. Like

High % Like

141

774

5.49

615

4.36

–

People Group /

NCP /

Night /

DBC News /

Low Acc. Age /

Low Avg. View /

Mid p95 Likes /

Male

JaPa

Late Night

Channel 24

Low View

Low Avg. Like

High % Reply

184

749

4.07

663

3.60

Romanized

Organization

JaPa

Early Morning /

DBC News /

Low Subscribe /

Low Avg. Like /

Low p95 Likes /

Bengali

Late Night

Independent TV

Low Video

Low Avg. View

Low % Like

Table 5: Condensed community-level behavioral profiles using dominant attributes from each feature group.

Comm.	Top Distinctive Attributes
C0	High Video (4.64), High View (4.55), Mid Video (2.77)
C1	Low Avg. View (2.61), Low Avg. Like (2.44), Mid p95 Likes (2.22)
C2	Low Avg. Like (2.75), Low Avg. View (2.59), Low p95 Likes (2.02)
C3	Mid Avg. Like (3.00), Mid Avg. View (2.74), Low % Like (1.93)
C4	Mid % Like (2.35), Mid % Reply (2.10), Mid p95 Likes (2.05)
C5	High Avg. View (2.61), High Avg. Like (2.53), Low % Like (2.03)

Table 6: Top three distinctive attributes of each community based on lift.

Figure 7: Community behavioral fingerprints based on column-wise standardized lift values.

Figure 8: Group-level differences between within-community and between-community similarity contributions.

Figure 9: Top 10 individual features based on within-community and between-community contribution differences.

Figure 10: Community-wise mean contributions of behavioral and contextual feature groups to within-community similarity.

Figure 11: Average influence spread of the seed-selection methods across communities for a seed budget of

k_{s}=3

Figure 12: Percentage difference between the greedy method and the strongest baseline across communities.

Comm.	IM $\cap$ B	IM $\cap$ WD	IM $\cap$ PR	B (%)	WD (%)	PR (%)
C0	0	0	0	0.00	0.00	0.00
C1	0	0	0	0.00	0.00	0.00
C2	1	0	0	33.33	0.00	0.00
C3	1	0	0	33.33	0.00	0.00
C4	0	0	0	0.00	0.00	0.00
C5	0	0	0	0.00	0.00	0.00

Table 7: Seed overlap between the greedy method and baseline strategies for

k_{s}=3

Figure 13: Representative diffusion dynamics of IM Greedy and Bully-Only in community C5.

CQ	Competency Question (Short)	Key Findings
CQ1	Identification of perpetrator users	Applying the threshold $\tau=3$ identifies $1{,}063$ users with at least three bullying-labeled comments or replies as perpetrators (Section 3.2.2 ).
CQ2	Behavioral characteristics of perpetrators	Community profiles reveal differences in user characteristics, video exposure, engagement response, language use, target identity, political context, temporal activity, and channel exposure (Section 4.3 ).
CQ3	Multi-scale network structure	The perpetrator similarity network contains $1{,}063$ nodes and $27{,}554$ edges, with an average degree of $51.84$ , clustering coefficient of $0.1939$ , and average weighted shortest-path distance of $3.4192$ in the largest connected component (Section 4.2 ).
CQ4	Community and sub-community structure	Louvain detection identifies six communities with a modularity of approximately $0.495$ . Hierarchical analysis further reveals $27$ sub-communities, exposing finer structural divisions within the parent communities (Section 4.2 ).
CQ5	Behavior–context association	Communities differ in behavioral profiles and contextual settings. Attributes associated with greater interaction volume do not necessarily exhibit higher bullying proportions (Section 4.4 ).
CQ6	Contributions to similarity and community separation	The Video group records the largest discriminative contribution, followed closely by the Impact and User Profile groups. These three groups also dominate the leading individual features, while their relative contributions vary among communities (Section 4.5 ).

Table 8: Summary of competency questions and corresponding findings.

Social Media Analysis

Social media analysis studies online posts, media, users, and interactions to understand communication, communities, and public behavior.

See all 222 papers

Oct 9, 2026

BenCoDiC: A Large-Scale Contextual Cyberbullying Corpus for Bengali Political News Discourse FULL VERSION

Cyberbullying is a growing concern in online communities, yet existing Bengali corpora remain limited in contextual information, annotation transparency, scale, accessibility, and support for multidimensional analysis, thereby hindering robust detection and comprehensive study of harasser behaviors and networks. To address these gaps, we introduce the Bengali Contextual and Diverse Cyberbullying (BenCoDiC) corpus, a publicly available contextual cyberbullying resource for Bengali political news discourse comprising 50,250 textual instances collected from 100 YouTube videos and their associated discussions. The corpus preserves caption--comment--reply relationships and is organized through a nine-dimensional star schema covering users, videos, textual content, context, time, engagement, political parties, target identities, and language categories. Five AI chatbots were initially evaluated for annotation but did not meet the predefined reliability criterion. Consequently, a two-tier human annotation process with expert verification was adopted, achieving Fleiss'

κ=0.86

and Krippendorff's

α=0.86

. Statistical analyses characterize the corpus, while transformer-based experiments show that incorporating conversational context yields modest but consistent improvements over isolated-text classification. BenCoDiC supports context-aware cyberbullying detection and perpetrator-oriented behavioral and network analysis. The corpus is available at https://github.com/bencodic/BenCoDiC-Corpus.

Md. Nesarul Hoque, Rudra Pratap Deb Nath, Umme Salma +2

Big Data, Information and Knowledge Engineering Lab, Department of Computer Science and Engineering, University of Chittagong, Chattogram-4331, Bangladesh · Department of Computer Science and Engineering, Gopalganj Science and Technology University, Gopalganj-8105, Bangladesh · Department of Computer Science and Engineering, Bangladesh University, Dhaka-1207, Bangladesh

Bengali NLP Social Media Analysis Hate Speech Detection

Ask PDF

May 26, 2026

Cyberbullying Governance on Social Media: A Unified Framework from Content Identification to Intervention

The proliferation of social media platforms and online communities has inadvertently catalyzed the spread of cyberbullying, hate speech, and other forms of online toxicity, making the effective governance of such harm a critical societal and computational challenge. While significant strides have been made in automating content moderation, existing research predominantly treats cyberbullying governance as passive, isolated detection at the post level. This reductionist view overlooks the continuous behavioral dynamics of users, the structural diffusion of toxic events, and the critical need for proactive mitigation. To bridge these gaps, this paper proposes a unified full-lifecycle governance framework that shifts the paradigm of cyberbullying governance from isolated static detection toward integrated, continuous, and proactive moderation. Drawing on cyberbullying research and adjacent fields, we systematically synthesize the state-of-the-art literature across four interconnected stages: (1) Content Identification, (2) User and Behavior Modeling, (3) Diffusion Dynamics and Early Warning, and (4) Intervention and Governance. Furthermore, we review available datasets and evaluation practices, and discuss emerging challenges including multimodality, explainability, algorithmic fairness, and the dual-use risks of generative AI, providing a roadmap for future research toward a safer and more resilient digital ecosystem.

Yiting Huang, Wenting Zhu, Zekun Wang +6

School of Cyberspace Security, Beijing University of Posts and Telecommunications, China · School of Cyberspace Security, Beijing University of Posts and Telecommunications,2026 China

Social Media Analysis Hate Speech Detection AI Governance

Ask PDF

Mar 2, 2026

Assessing Crime Disclosure Patterns in a Large-Scale Cybercrime Forum

Cybercrime forums play a central role in the cybercrime ecosystem, serving as hubs for the exchange of illicit goods, services, and knowledge. Previous studies have explored the market and social structures of these forums, but less is known about the behavioral dynamics of users, particularly regarding participants' disclosure of criminal activity. This study provides the first large-scale assessment of crime disclosure patterns in a major cybercrime forum, analysing over 3.5 million posts from nearly 300k users. Using a three-level classification scheme (benign, grey, and crime) and a scalable labelling pipeline powered by large language models (LLMs), we measure the level of crime disclosure present in initial posts, analyse how participants switch between levels, and assess how crime disclosure behavior relates to private communications. Our results show that crime disclosure is relatively normative: one quarter of initial posts include explicit crime-related content, and more than one third of users disclose criminal activity at least once in their initial posts. At the same time, most participants show restraint, with over two-thirds posting only benign or grey content and typically escalating disclosure gradually. Grey initial posts are particularly prominent, indicating that many users avoid overt statements and instead anchor their activity in ambiguous content. The study highlights the value of LLM-based text classification and Markov chain modelling for capturing crime disclosure patterns, offering insights for law enforcement efforts aimed at distinguishing benign, grey, and criminal content in cybercrime forums.

Raphael Hoheisel, Tom Meurs, Jai Wientjes +3

University of Twente, Netherlands · Dutch Police, Netherlands · Université de Montréal, Montréal Complexity Science Hub, Austria

Cybersecurity

Ask PDF

Abstract

Social Media Analysis

Explore similar work

BenCoDiC: A Large-Scale Contextual Cyberbullying Corpus for Bengali Political News Discourse FULL VERSION

Cyberbullying Governance on Social Media: A Unified Framework from Content Identification to Intervention

Assessing Crime Disclosure Patterns in a Large-Scale Cybercrime Forum