• Conferenza GARR 2018

     

     

       GARR Conference is the annual meeting of users, operators and managers of the Italian national education and research network aimed at sharing experiences and comments on the use of the network as a tool for research, training and culture, in different contexts and disciplines.
   

     

     

Temi della conferenza    

Temi della conferenza    

The core themes of this editions are data, artificial intelligence and technology transfer both towards the enterprises and single persons. We will talk about open data and services, cybersecurity, industry 4.0 and its relations with research and innovation. We will discuss how education should be rethought in order to keep the pace with the continuous evolution of ICTs, also with the help of these very same technologies.   

Programme


Corsi di formazione   

Corsi di formazione   

On 1-2 October there will be several training opportunities for GARR user community. These courses span from classes dedicated to trainers (Moodle: how to manage a course e How to create and disseminate educational material online), to courses on Public Speaking, on cybersecurity (PenTest & Rooting) and on networking (Software Defined Network). For registrations to these training courses, go to the Learning GARR platform.    

Programme

3 ottobre 2018

Francesco La Rosa

Università di Messina
https://www.unime.it/

Soluzioni di Deep Learning per la Cyber Security

Laureato in Ingegneria Elettronica, ha conseguito un dottorato di ricerca in Computer Science presso l’Università degli Studi di Messina. È coautore di decine di articoli pubblicati su atti di convegno e riviste internazionali nell'ambito della visione artificiale, videosorveglianza, pattern recognition e reti neurali. Negli ultimi anni ha ricoperto ruoli di responsabilità c/o il CIAM (Centro Informatico Ateneo di Messina), Università degli Studi di Messina. Attualmente sta svolgendo attività di ricerca nei settori del Machine Learning e del Deep Learning con applicazioni alla videosorveglianza e alla Cyber Security.

Graduated in Electronic Engineering, he obtained a Ph.D. in Computer Science at the University of Messina. He is co-author of dozens of articles published on conference proceedings and international journals in the field of artificial vision, video surveillance, pattern recognition and neural networks. In recent years he has held positions of responsibility at CIAM (University Information Center of Messina), University of Messina. He is currently carrying out research activities in the fields of Machine Learning and Deep Learning with applications for video surveillance and Cyber Security.

SESSIONE 4. CYBERSECURITY e AI

Soluzioni di Deep Learning per la Cyber Security

Con la crescente integrazione tra Internet e i social media è radicalmente cambiato il modo di apprendere, comunicare e lavorare, esponendo chi fa uso di queste tecnologie a crescenti minacce/attacchi alla propria sicurezza. L’identificazione dei vari attacchi di rete, in particolare di quelli mai visti in precedenza, è un problema da risolvere con urgenza. Un componente fondamentale di un’infrastruttura di cyber security è il sistema di Network Intrusion Detection (NIDS). Un NIDS viene usato per identificare, analizzando il traffico di rete su nodi chiave, attività malevole volte a violare la confidenzialità, l’integrità e la disponibilità dei dati e dei sistemi. Molti tra i NIDS più moderni fanno uso di tecniche di Machine Learning (ML) o Deep Learning (DL) per la loro capacità di adattamento ad attacchi di rete sconosciuti (zero-day). In questo articolo proponiamo un NIDS basato su una deep neural network già adottata con risultati notevoli in ambiti come quelli della Computer Vision e del Natural Language Processing. I risultati sperimentali mostrano che la nostra soluzione supera approcci analoghi, presenti in letteratura, in termini di accuratezza (Acc), detection rate (DR) e False Acceptance Rate (FAR).


Deep Learning solutions for the cybersecurity

With the growing integration between Internet and social media, the way we learn, communicate and work has radically changed, exposing the users of these technologies to increasing attacks on their security. The identification of various network attacks, in particular those previously unseen, is a problem to be solved urgently. A core component of a computer security infrastructure is the Network Intrusion Detection (NIDS) system. An NIDS is used to identify, by analyzing network traffic on key nodes, malicious activity aimed at violating the confidentiality, integrity and availability of data and systems. Many of the most modern NIDS make use of Machine Learning (ML) or Deep Learning (DL) techniques for their ability to adapt to unknown network attacks (zero-day). In this article we propose a NIDS based on a deep neural network already adopted with remarkable results in areas such as Computer Vision and Natural Language Processing. The experimental results show that our solution exceeds similar approaches (present in the literature) in terms of accuracy (Acc), detection rate (DR) and False Acceptance Rate (FAR).

 


Sponsor

DELL EMC
Palo Alto Networks
INTEL

juniper networks
maticmind