Image-Centric Social Discovery Using Neural Network under Anonymity Constraint

A common part of social networks is sharing images. Storing and processing these images to provide engaging services to customers is resource intensive and social networks often rely on cloud services to provide these resources. This creates a potential security risk for the users of the social network platform. Simply encrypting the images before uploading them to the cloud service does solve part of this problem but it also reduces a lot of the functionality that the social network and its users desire in terms of image content analysis. Kazi Wasif Ahmed, Mohammad Zahidul Hasan and Noman Mohammed from the University of Manitoba propose a method for mitigating this risk. By extracting features from the images using machine-leaning analysis before uploading them this functionality is preserved. The cloud service is then able to perform searching on anonymized profiles, retaining the required functionality while reducing the risk to the service users. This method provides similar performance to searches that do not provide the protection. The research provides insight into a way by which greater safety can be provided to users of social network without sacrificing features.



Ahmed, K. W., Hasan, M. Z., & Mohammed, N. (2017, April). Image-Centric Social Discovery Using Neural Network under Anonymity Constraint. In Cloud Engineering (IC2E), 2017 IEEE International Conference on (pp. 238-244). IEEE.