Traditional access control systems, reliant on static credentials like PINs or RFID cards, are ill-suited for the complex social dynamics of modern smart buildings. These environments involve a diverse and transient set of individuals whose interactions are unpredictable and cannot be managed by rigid, pre- defined rules. This paper confronts this challenge by introducing a proof-of-concept (PoC) for an adaptive access control system that intelligently manages trust in these dynamic scenarios. Our framework integrates four key technologies, chosen specifically to address the limitations of static systems. To understand the crucial context of a visit, a Natural Language Processing (NLP) module interprets a visitor's spoken purpose, providing structured data on their intent. To create an evolving security posture, this data informs Dynamic Trust and Anomaly Detection modules, which maintain a persistent trust score for each visitor based on their behavior over time. The system's decision-making core translates this contextual data into a quantifiable security response. The Dynamic Policy Module uses the NLP context and real-time trust score to dynamically set the voting threshold (t) required for entry. To ensure authorization is both secure and democratic, a Cryptographic Voting Module uses Shamir's Secret Sharing (SSS). This choice facilitates decentralized decision- making among residents, eliminating single points of failure and reliance on a single trusted authority. Our results provide quantifiable evidence of its effectiveness: in response to high-risk scenarios, the system automatically hardened its security posture, increasing the required voting threshold to a 75% supermajority. This enhanced security is achieved with acceptable performance costs for real-time interaction; the core intelligence modules (NLP, Trust, and Policy) add a consistent overhead of only 10-50 ms per request, with the cryptographic core maintaining real- time viability for communities of up to 120 residents.
Access Control, Threshold Cryptography, Natural Language Processing, Trust Management, Anomaly Detection
© IACyC is conducted under the administration of the CyberMACS Erasmus-Mundus Masters Programme. cybermacs.org