Hacking smart machines with smarter ones

Luigi V. Mancini

14Jan
14:00 14th January 2019 ( Hilary Term 2019 )
Tony Hoare Room, Robert Hooke Building

Thanks to Machine Learning (ML) techniques, computers learn to recognize profiles and trends, make decisions or react autonomously to dynamic environments. Although ML's algorithms are known, the datasets used to train them are usually not public and, in fact, can be protected by privacy laws, or can be kept as trade secrets.

This presentation focuses on ML classifiers and how sensitive information associated with the training dataset can be reconstructed through legitimate interactions with them. The Generative Adversarial Networks (GANs) model is presented and it illustrates how GAN neural networks can be used to attack other automatic learning systems. In particular, it is shown that it is possible to deduce sensitive information from ML classifiers. This information leakage can be exploited, for example, by a competitor to build more effective classifiers, to acquire sensitive user data, or to acquire trade secrets by interacting with an apparatus, potentially violating its intellectual property rights.

Speaker bio

Luigi V. Mancini is full professor at the University of Rome “La Sapienza” (Italy), and the director (Presidente) of the local Master of Science program (Laurea magistrale) in Cybersecurity.

The current research interests of Luigi V. Mancini include: network and information security, and user privacy. Luigi V. Mancini published more than 110 scientific papers in international conferences and journals. He has served on the program committees of several international conferences, among which: ACM Conference on Computer and Communication Security, ACM Symposium on Access Control Models and Technology, European Symposium on Research in Computer Security, and Financial Cryptography and Data Security Conference.

He is the founder of several Master degree programs in Information and Network Security at the University of Rome "La Sapienza". He participated in numerous national and international research projects in the area of security.

Luigi V. Mancini received the PhD degree in Computer Science from the University of Newcastle, UK, in 1989.

Citations: 6365 (December 2018). H-index: 40, based on Google Scholar

Web: http://www.di.uniroma1.it/mancini