Artificial intelligence and machine learning approaches for secure image recognition in IoT networks

The extremely fast expansion of Internet-of-Things (IoT) networks has generated high security concerns, in particular for the integrity and privacy of image data. Private image recognition is essential to ensuring that image data captured by IoT devices (surveillance cameras, sensors, etc.) is reliable and secure by preventing unauthorized access or tampering of this data. In this paper, we propose a new methodology that uses the power of Artificial Intelligence (AI) and Machine Learning (ML) to improve security of image recognition systems in the IoT environment. Concretely, convolutional neural network (CNN), image encryption and adversarial machine learning are involved in the proposed image recognition strategy for security purpose, which can examine against image tampering and unauthorized access and so on. Experimental results show that the proposed method is capable of effectively recognizing image and ensuring data security and integrity, and gains great superiority over those of other existing methods in security and recognition rate. The proposed framework provides a suitable solution to protect the image-based data within IoT platforms, and thus, to have more secure and reliable IoT platforms.