Lightweight CNN-based edge computing with hybrid chaotic encryption for secure and efficient image transmission in IoT networks

While IoT Devices are used in Governmental organizations to send data to Cloud services, there are several Issues with conventional communication methods used to connect devices.1) The secure transmission of confidential Images is not guaranteed.2) Limitations of Memory, Computation, and Bandwidth are present on IoT devices. This Paper presents the development of a framework for the secure transmission of confidential images in real-time in Indian Government IoT Environments using a CNN-assisted Selective Encryption Framework. An Edge-based CNN, which consumes minimal processing power to classify images into either Sensitive or Non-sensitive, is employed to perform automatic classification of digital images. Once classified by the CNN, the Sensitive images will be transmitted using a Chaotic Hybrid encryption scheme in conjunction with the SHA-512 algorithm, thus ensuring Confidentiality and Integrity. Conversely, non-sensitive images will use no Encryption, allowing for a reduction in both processing overhead and service Latency. As a result, the experimental results show that the classification results achieved high accuracy (Training: 98.6%, Validation: 97.2%, Testing: 96.8%) and that the Encryption Algorithms employed in this framework demonstrated state-of-the-art cryptographic security through the analysis of Ciphertext Entropy (7.9972), Correlation (Minimal) and NPCR/UACI values demonstrating high resistance to statistical and differential attacks. Through comparisons made against existing encryption schemes, the Chaotic Hybrid Encryption approach used in this Paper demonstrated an approximate 6.23 times faster throughput than the existing schemes, and a 6.22 times reduction in overall computational Latency, thus presenting new ways of transmitting secure messages through Government IoT Communication in real-time.