A Multilayered Trusted and Secure Surveillance with Frame Encoding, ML, and Hashgraph
Ayush Verma, Tanuj Chandela, Geetanjali Rathee, Abhinav Tomar · 2024
In today's interconnected society, ensuring the safety and security of the public is crucial for protecting individuals from threats posed by criminals. Equally important is safeguarding individuals' privacy, which requires the development of a robust mechanism capable of withstanding attacks while remaining efficient. Surveillance cameras, such as Closed-Circuit Television (CCTV), have become essential tools in this endeavour. These cameras capture extensive visual data, often processed and stored on remote, centralized cloud servers. However, these methods have several shortcomings and are vulnerable to numerous security and privacy breaches. Various solutions have been proposed to address these challenges, each focusing solely on specific aspects of the system. However, none of them have comprehensively addressed the complete mechanism and discussed its resilience against various cyber threats. This paper proposes a secure and trusted decentralized surveillance mechanism utilizing hashgraph for faster, scalable, and secure consensus. The framework aims to defend against attacks such as DDoS and Sybil Attacks and ensures tamperproof logging of all events securely. For rapid identification, we employ I-frame generation from camera footage and a two-stage detection process: one at the camera end using the MobileNetV3-small model to forward necessary keyframes, and another on a private cloud-based analyzer. To safeguard individuals' privacy and ensure secure I-frame transmission while preventing attacks such as Known-plaintext and Replay attacks, we implement an XORing mechanism alongside AES (Advanced Encryption Standard) for I-frames.