16 January. 2021
Every sensor and embedded device is prone to error failure which can be caused by natural causes such as environmental effects, battery discharging or by malicious invasion to the network. A fault in a node can decrease network performance and/or in the worst case scenario, dissolution of the network. Diagnosing faults in the network at an early stage can decrease the possibilities of tear down the network. Fault diagnosis can help identify the nature of the error; whether the error is a result of malicious intervention or of natural causes. The goal of the current project is to create faults so that to train a diagnosis tool. The tool will evaluate a set of data that will be considered crucial to identify the presence of fault or malicious intervention in the network. Data gathered in a controlled environment will be profiled as normal behavior, that is, behavior with no faults present in the network. Data will also be taken in simulations where faults, or attacks are present. Both benign and malicious data will be used to establish boundaries that will identify the presence of malicious attack of failure. The recognition and classification of the activity will be based on one or more techniques from Statistical Analysis, Machine Learning and Computational Intelligence.
Computer Networks, Network protocols, Basic Programming, basic understanding of AI/ML
To create new intrusion detection techniques, or the extensive evaluation of existing techniques.
Expected deliverables: Final Report, Simulation Scripts