Home/Problem Statements/SIH26189

SIH26189SoftwareBlockchain & Cybersecurity

₹1,50,000 Prize (₹75k Joint)

•

Deadline: 20 September 2026

•

Submitted Ideas: 0/500

AI-Powered Criminal Network Analysis System

Ministry / Organization: Ministry of Home Affairs

Department: National Crime Records Bureau (NCRB), Women Safety Division

SIH Official Portal

Technical Complexity

8/10

Dataset Accessibility

7/10

36h Hackathon Viability

9/10

Step 0 • Zero-Jargon Layman Summary

“Law enforcement agencies collect massive amounts of crime data across different files and formats, making it hard to connect the dots between criminals. We need to build an AI system that automatically reads this scattered data, maps out hidden relationships, identifies ringleaders, and visualizes criminal syndicates for investigators.”

The DecodeX PS Decoder™

Deconstructed Problem Blueprint

Step 1

WHAT TO BUILD?

An end-to-end AI-powered Graph Analytics and Intelligence Platform that ingests unstructured and structured law enforcement data, performs Named Entity Recognition (NER), builds an interconnected knowledge graph, and surfaces criminal networks.

Step 2

WHO IS IT FOR?

The problem belongs to the Ministry of Home Affairs (National Crime Records Bureau, Women Safety Division). The users are crime investigators, intelligence analysts, and law enforcement officers.

Step 3

WHY DOES IT EXIST?

Current crime data is siloed across different police stations, formats (PDFs, FIR text, audio, logs), and databases. Investigators rely heavily on slow, manual tracking which fails to uncover sophisticated, distributed criminal networks operating across state and national borders.

Step 4

INPUT → PROCESS → OUTPUT ARCHITECTURE PIPELINE

01 • Inputs & Telemetry

FIR PDFs, police reports, Call Detail Records (CDRs), financial transaction logs, surveillance text, social media dumps, and criminal history databases.

02 • Processing & AI Core

NLP-based Named Entity Extraction (people, locations, accounts), relationship extraction, vector embedding generation, community detection graph algorithms (e.g., Louvain, PageRank), and suspicious pattern matching.

03 • Actionable Deliverables

Interactive Knowledge Graph UI with node-link diagrams, risk-scored suspect profiles, automated link analysis reports, and real-time alerts for high-risk associations.

Step 5

CONSTRAINTS

Must handle data privacy and anonymization rules compliant with Indian legal frameworks (e.g., BNS, BSA)
Must support ingestion of multi-lingual text and regional Indian languages/scripts found in FIRs
Low-latency graph queries even when traversing 100k+ nodes
System must function in secure, air-gapped environments with restricted internet access

Step 6

JURY METRICS

Entity extraction and linking precision and recall accuracy
Graph query and visualization rendering latency (<2 seconds for 10k nodes)
User interface simplicity and intuitive navigation for non-technical field investigators
Accuracy of identifying true key influencers using centrality algorithms

Step 7

CORE BOTTLENECK

Extracting accurate relationships and entities from highly unstructured, noisy, multi-lingual FIR texts and merging duplicate entities (e.g., matching 'Rajesh Kumar', 'R. Kumar', and 'Raju') into a single identity node.

Why You Should Choose This PS (Pros)

Extremely high national security and social impact for the Ministry of Home Affairs
Clear, tangible deliverables (Graph UI + NER backend + Intelligence Reports)
Allows showcasing cutting-edge AI architectures like Graph Neural Networks (GNNs) and LLM-powered RAG
Strong potential for real-world deployment and prototype piloting with NCRB

Challenges & Risks to Watch (Cons)

Lack of authentic, real-world classified datasets requires building realistic, domain-accurate synthetic data pipelines
Complex graph visualization rendering can lag if not optimized correctly for web browsers
Crowded problem domain if teams build standard CRUD apps without deep graph analytical insights

Hackathon Blueprint

Recommended Tech Stack & Grand Finale Features

Frontend & Mobile

React.jsTailwind CSSVis.js or Cytoscape.js for interactive graph visualization

Backend & APIs

FastAPI / PythonCelery for asynchronous background data processing tasks

AI / ML & Models

spaCy / IndicBERT for multilingual NERNetworkX / Neo4j Graph Data Science (GDS) algorithmsSentence-Transformers for semantic entity matching

Database & Cloud

Neo4j or PostgreSQL with Apache AGE / pgvector for graph and relational queriesRedis for real-time caching

Killer Features to Impress the SIH Jury

AI-Powered Alias & Entity Deduplication: Automatically merges varying spellings and aliases of suspects across documents into a single master node.

Automated 'Kingpin' & Influencer Identification: Uses PageRank and Betweenness Centrality metrics to instantly highlight who pulls the strings behind the scenes.

Time-Slice Criminal Timeline Replay: A slider interface that lets investigators playback the evolution of a crime network chronologically over weeks or months.

Natural Language Investigative Query Box: Allows officers to type plain English queries like 'Show all financial transactions between suspect A and unknown shell companies in the last 30 days'.

View Official SIH 2026 Portal Raw Text

▼

•Background Modern criminal activities are increasingly organized and interconnected. Criminals often operate through networks involving associates, intermediaries, financial channels, communication links,locations, and events. Law enforcement agencies collect large volumes of data from sources such as:

•FIRs and police reports

•Call Detail Records (CDRs)

•Financial transaction records

•Surveillance reports

•Social media intelligence

•Criminal history databases

•Intelligence agency reports Despite having access to this information, investigators frequently face challenges in identifying hidden relationships among suspects because the data is fragmented, unstructured, and distributed across multiple systems. Manual analysis can be slow, labor-intensive, and prone to missing critical connections.With advances in Artificial Intelligence (AI), Machine Learning (ML),Natural Language Processing (NLP), and Graph Analytics, it is now possible to automatically discover relationships, detect patterns, and generate insights that can assist investigators in understanding criminal networks more effectively.

•Description The objective is to develop an AI-powered system that can analyze large volumes of criminal and intelligence-related data to uncover hidden networks and relationships among individuals, organizations, locations,and events.

The system should:

•Collect and process data from multiple sources.

•Extract important entities such as people, locations, vehicles, phone numbers, and organizations.

•Build relationship maps showing how different entities are connected.

•Identify key individuals who play influential roles within criminal networks.

•Detect suspicious patterns and unusual activities.

•Assist investigators by providing visual and analytical insights.

•Expected Solution Develop an AI-powered system that automatically analyzes structured and unstructured crime-related data to uncover criminal networks,identify key influencers, detect suspicious patterns, and provide actionable intelligence for investigators.