A Secure Context-Aware Middleware for Computation Offloading in Untrustworthy, Open, and Dynamic Edge EnvironmentsMaster Thesis
3 February 2022, by Kevin Röbert

Photo: midjourney
Abstract:
Computation offloading systems promise to remove the boundaries of local computing capacity and provide access to almost infinite external resources. These resources are often backed by the cloud, yet many unused resources exist in the edge, particularly at residential and institutional locations. However, offloading on edge resources does not come without a price – the heterogeneity of hardware, network connectivity, reliability, and decentralized nature pose new challenges. Especially in open and untrusted networks like the Internet, network barriers such as NATed firewalls, dynamic IP addresses, roaming, and non-public nodes often prevent access to these resources. Thus, research on distributed computing frequently abstracts from these real-world conditions and either assume a reliable network without barriers and errors or declares them out-of-scope. This makes statements about the approaches on real networks difficult. This thesis designs, implements, and evaluates a middleware for computation offloading that takes the real-world constraints into account and gives appropriate solutions to unleash the full potential of the edge – allowing us to fulfill the promise of ubiquitous computing and evaluating under present conditions by running several realworld simulations.
Supervised by:
Prof. Dr. Janick Edinger, Heiko Bornholdt