Anti-fragile internet with autonomous swarm networks

• Main Author: Lua, Rui Ping
• Other Authors: Ng Wee Keong
• Format: Theses and Dissertations
• Language: English
• Published: 2015
• Subjects: DRNTU::Engineering::Computer science and engineering::Data
• Online Access: https://hdl.handle.net/10356/62139
• Institution: Nanyang Technological University

Internet services and traffic is growing at an exponential rate. They are however vulnerable to flash crowds and Distributed Denial of Service attacks. Existing techniques are difficult to scale and have limited effectiveness. Most of them addresses specific attacks and do not provide wider coverage. We explore concepts of anti-fragility, autonomic and swarm computing to address these problems. We propose a decentralized and iterative overlay structure to reduce disruptions and risks in large networks. An autonomic system features (1) Self-configuration, (2) Self-healing, (3) Self-optimization and (4) Self-protection. For each of the features, we propose appropriate mechanisms. We show how each component is integrated to addresses simple and sculpted attacks. Fast flux session binding allows clients to contact gateway nodes in overlay networks. Auto-structure describes how the network builds itself iteratively. Auto-sensing estimates global traffic flows effectively. Auto-resistance deploys traffic scrubbers in real-time to mitigate illegitimate traffic. Auto-flow optimization increases network throughput through route selection. Our system filters traffic that does not conform to expectations of end point servers. Application level attacks will trigger auto deployment of filters. Meanwhile, self-management strategies constantly optimize and heal the underlying network. Simulations were ran to demonstrate effectiveness of the above features. These were then integrated into a prototype. This prototype was deployed to a large number of computers to demonstrate how each feature performs. Results from our deployment has shown that we can mitigate application level attacks effectively. We are also able to perform real-time optimization of our network by varying the number of active nodes and their respective pools. We show how swarm algorithms such as IWD can be used to perform distributed traffic management. This allows us to allocate traffic effectively and increases survivability of the network.