Modern Intrusion Detection Systems (IDSs) rely on machine learning for detecting and defending cyber-attacks in information technology (IT) networks. However, the introduction of such systems has introduced an additional attack dimension; the trained IDS models may also be subject to attacks.
The act of deploying attacks towards machine learning based systems is known as Adversarial Machine Learning (AML) . The aim is to exploit the weaknesses of the pretrained model which has “blind spots” between data points it has seen during training. More specifically, by automatically introducing slight perturbations to the unseen data points the model may cross a decision boundary and classify the data as a different class. As a result, the model’s effectiveness can be reduced as it is presented with unseen data points that it cannot associate target values to, subsequently increasing the number of misclassifications.
Adversarial machine learning attacks and automated detection of these attacks in computer networks will be investigated in this project.
 Shirazi, Shirin Haji Amin. A Survey on Adversarial Machine Learning.
- Identifying the state of the art for adversarial attacks to machine learning models.
- Investigating the impact of adversarial attacks on network security.
- Identifying and implementing a cybersecurity defense strategy for networks that can automatically detect adversarial exploiting the vulnerability of machine-learning-based models.
- Review possible adversarial attacks to machine learning models in cybersecurity.
- Proposed and implement a solution for automatically detecting adversarial attacks against machine-learning-based algorithms in cybersecurity.
Skills and experience
Basic knowledge of computer networks and machine learning.
Contact the supervisor for more information.