Combining Concolic Testing with Machine Learning to Find Software Vulnerabilities in the Internet of Things

Concolic testing is a software verification technique that has been successfully applied to find subtle bugs in embedded software. In particular, it relies on efficient symbolic execution engines to produce program inputs, which can be used to concretely execute the program under analysis with the goal of achieving high code coverage. Machine learning techniques have also merged as an efficient approach to predict properties of the program or to identify regions of the state-space to be explored for some particular property. Given that Internet of Things (IoT) is now present in all technology sections, allowing different systems to connect and exchange data, the identification of software vulnerabilities in IoT devices has become a major concern in large IT organisations. This PhD research is concerned with identifying software vulnerabilities by combining concolic testing with machine learning techniques to prevent unauthorised access to the IoT devices. In particular, the main objectives of this PhD research are: (1) analyse and develop a deeper understanding of software security as a whole to capture main properties of interest to a secure network in IoT; (2) understand all possible cyber threats/attacks that IoT devices can face in order to shield the network from malicious attacks, thus protecting the data flowing through the network; (3) propose an efficient method to identify software vulnerabilities using concolic testing and machine learning techniques, in order to make IoT devices less susceptible to the cyber threats/attacks; (4) apply this verification method to a large number of open source applications that can benefit from a rigorous software security analysis.