MACHINE LEARNING BASED HEALTHCARE SYSTEM FOR INVESTIGATING THE.ASSOCIATION BETWEEN DEPRESSION AND QUALITY OF LIFE
Abstract : Machine learning (ML) algorithms are nowadays widely adopted in different contexts to perform autonomous decisions and predictions. Due to the high volume of data shared in the recent years, ML algorithms are more accurate and reliable since training and testing phases are more precise. An important concept to analyze when defining ML algorithms concerns adversarial machine learning attacks. These attacks aim to create manipulated datasets to mislead ML algorithm decisions. In this work, we propose new approaches able to detect and mitigate malicious adversarial machine learning attacks against a ML system. In particular, we investigate the Carlini-Wagner (CW), the fast gradient sign method (FGSM) and the Jacobian based saliency map (JSMA) attacksThe aim of this work is to exploit detection algorithms as countermeasures to these attacks. Initially, we performed some tests by using canonical ML algorithms with a hyperparameters optimization to improve metrics.Then, we adopt original reliable AI algorithms, either based on eXplainable AI (Logic Learning Machine) or Support Vector Data Description (SVDD). The obtained results show how the classical algorithms may fail to identify an adversarial attack, while the reliable AI methodologies are more prone to correctly detect a possible adversarial machine learning attack. The evaluation of the proposed methodology was carried out in terms of good balance between FPR and FNR on real world application datasets
? Until now, this attack is considered as the most powerful among the existing attacks. Obviously, since the generation of malicious adversarial data is very accurate, computational times can be very long.
? ML and artificial intelligence (AI) algorithms have been applied to many and different contexts in recent years, from the healthcare world to intrusion detection systems in the field of IoT security.
? We chose the mentioned algorithms as they are among the most used ones for binary classification problems in recent machine learning literature .
? In the platooning dataset, on the other hand, a strong superposition of points between the two classes makes the detection a hard task. Finally, in the RUL estimation original problem, the healthy and fault classes are quite well separated and we will investigate how the different proposed attacks impact on this base performance
? The proposed activities are tested and evaluated on three different datasets: the first one is focused on network security (in particular, a DNS tunneling communication), the second one focused on vehicle platooning and the third one is a benchmark in predictive maintenance, consisting in Remaining Useful Life (RUL) estimation.
? Finally, in the RUL estimation original problem, the healthy and fault classes are quite well separated and we will investigate how the different proposed attacks impact on this base performance.
? Machine learning (ML) has become an increasingly used technology in every aspect of our lives.
? Then offers a broad overview of the most widely used and efficient methodologies for dealing with adversary attacks in AI fields.
? The perturbations performed by the adversarial machine learning attacks aim to be minimal to fool the model without an obvious change in the data used.
|