Semantic-based network behavior analysis and its utilizations

Abstract

Efficiently managing and controlling the network is one of the biggest problems in this present living world. In order to understand low-level network messages, we need a deep understanding of network service-specific knowledge. Moreover, due to the explosive growth of mobile and cloud services, the number of network services has also grown significantly, making existing problems even more difficult.

In this dissertation, we argue that understanding the semantics of network traffic leads to remedies for more efficient and effective network service management. To substantiate our claim, we present the design and implementation of two network systems. First, we show Lumos that improves IoT interoperability by leveraging Android apps that control Internet-of-Things (IoT) devices and learn from their behavior. The semantic identification methodology for low-level network messages devised by Lumos can overcome various problems in mobile traffic management, not only IoT. Next, we present Evanesca, a model-based DeFi analysis framework to uncover the abnormal behavior that has recently plagued DeFi services. Furthermore, its approach can be applied not only to DeFi services but also to various decentralized network services such as game, auction, NFT market, and decentralized court, thus it is a new behavior analysis methodology specialized in blockchain ecosystems.