Shirisha Reddy V , Trisha Anbu Kumar , Chaithra S, Praveena K N "Privacy-Preserving, AI-Driven Crowdsensing Framework for Closed-Loop Civic Infrastructure Management"

Paper_id: 56_7

Abstract

Rapid urbanisation is making the conventional civic grievance redressal systems less efficient. This depends heavily on manual sort, department wise routing, and storage of personally identifiable information (PII). These deficiencies undermine both resolution efficiency and citizen participation rates. This paper describes the design, implementation, and evaluation of Spotit.Fixit, which is a multi-tenant, crowdsourced e-governance platform engineered to address these gaps through three integrated mechanisms. First, a transformer-informed Natural Language Processing (NLP) classification pipeline automatically routes unstructured citizen complaints to the appropriate municipal authority, eliminating first-tier human dispatch. Second, a cryptographic identity masking protocol decouples national identification numbers (Aadhaar) from transactional data using one-way hashing, generating audit-traceable anonymity masks that satisfy Role-Based Access Control (RBAC) constraints. Third, a community driven bidirectional verification protocol requires municipal administrators to submit geotagged photographic proof of repair; case closure is only permitted once a 30% community confirmation threshold is reached via tokenised, email based upvoter verification, mathematically preventing unilateral ticket falsification. The framework additionally incorporates a real-time geospatial cross-validation and cryptographic image hashing layer that detects photographic evidence submitted from coordinates inconsistent with the reported incident site, flagging potential fraud and enabling Aadhaar linked legal accountability. Evaluation against a live demonstration corpus of 95 simulated civic complaints produced an aggregate NLP routing accuracy of 92.6% a reduction in mean triage latency from 48–72 hours to under two seconds , and a verified elimination of fraudulent ticket closures. These results demonstrate the system's suitability as a deployable civic accountability infrastructure for data-dense metropolitan environments.