Luca Didaci is an associate professor at the Department of Electrical and Electronic Engineering,  University of Cagliari. At the same university, he obtained a degree in Electronic Engineering in 2001 and a PhD in Electronic and Computer Engineering in 2005. Since 2006, prof. Didaci is a researcher for SSD ING-INF/05, initially at the Department of Pedagogical and Philosophical Sciences, and since September 2011 at the Department of Electrical and Electronic Engineering.

His research activity  takes place in the field of pattern recognition and its applications, and covers the following topics.

1) Computer security. Intrusion Detection Systems (IDS) and antiviruses aim to recognize 'malicious' software and behavior on computers and computer networks. However, traditional systems have a high false alarm rate and a rigidity that prevents the recognition of new attacks. Dr. Didaci worked on the study of new methods based on Machine Learning and Multiple Classifier Systems (MCS) characterized by an increase in accuracy and a better balance between generalization capacity and false alarm generation. As part of the CyberROAD EU project, he contributed to defining a methodology for creating research paths on IT security issues.

2) Security and Biometric Recognition. Biometric systems aim to use physiological or behavioral characteristics for individual recognition. In this context, multimodal systems based on the automatic recognition of faces and fingerprints and systems based on the analysis of brain signals have been studied. In particular, the study of brain signals highlights the discriminating power of functional connectivity measures obtained from the EEG signal for biometric identification purposes. Finally, the security properties of machine learning algorithms in the biometric field against specifically targeted attacks and the related countermeasures were studied.

3) Study of Multiple Classifier Systems. Multiple Classifier Systems (MCS) are a state-of-the-art approach for the design of classification algorithms, and allow to overcome several limitations of the traditional approach, based on the use of a single algorithm. The research activity focused on two strands. The first is related to the dynamic selection methods of the classifier, in which metrics are used that assume a non-isotropic feature space and adaptive metrics for estimating the 'degree of competence' (local accuracy) of the classifier. The second is related to the formulation of new methodologies for creating classification systems and to the study of diversity as a useful measure for their creation.

4) Semi-supervised learning. Semi-supervised learning aims at the joint use of data whose class of membership is known and data of unknown class, whereas traditionally the classification algorithms were based only on data of known class. This approach would make it possible to produce classification systems that “get better with use”. These methods have been extended both to multiple classifier systems and to innovative systems in the field of biometric recognition. In particular, in the biometric field the time-variance (due to aging phenomena or temporary changes to biometrics) and the considerable variability of biometric traits make the use of semi-supervised methods extremely appropriate.

5) Analysis and characterization of EEG signals using machine learning techniques.

Prof. Didaci is a member of the research laboratory on pattern recognition and its applications (PRA Lab, https://sites.unica.it/pralab/), of the Italian Group of Pattern Recognition Researchers (Italian Section of the International Association for Pattern Recognition ) and the IEEE (Institute of Electrical and Electronics Engineers).

[Last update: September, 2023]