Multi-view convolutional neural networks for cybersecurity applications

Cybersecurity therefore is of fundamental importance to defend against malicious actors who regularly target ubiquitous technologies like Android OS and web applications. In theory, defensive solutions using artificial intelligence, machine learning and neural networks can offer a significant degree of automation.

However, false positive rates of these systems remain a problem, whilst there are issues translating state-of-the-art lab research into the real-world, not least the difficulty in evaluation with a scarcity of representative data that is often proprietary and highly-sensitive. To help solve these challenges, this thesis presents three multi-view convolutional neural networks to advance the fields of Android malware detection and web application security.<br><br><br>