Sentient Networks: A New Dimension in Network Capability

Abstract

As computer networks evolve to the next generation, they offer flexible infrastructures for communication as well as high vulnerability. In order to sustain the reliability, to prevent misuse, and to provide radically new services, these networks need new capabilities in non-traditional domains of operation such as perception and consciousness. Therefore, we propose Sentient Networks, as an early approach towards realizing them. There are several applications for such networking capability. For one, sentient networks can enable packet detection in a blackbox manner, detect encryption of data transferred, detecting and classifying data types such as voice, video, and data. Some of the applications of such capabilities, for example, include the use of emotion content of the voice packets through a network and classifying these packets into normal, moderately panicked, and extremely panicked. Such classifications can be used to provide better QoS schemes for panicked callers. We obtained a packet detection accuracy of about 65-70% for data packets and about 98% accuracy for detecting encrypted TCP packets. The emotion-aware QoS provision mechanism provided approximately 60% improvement in delay performance for voice packets generated by panicked sources.