Computer Science

Computing over encrypted data
  • Traditional cryptographic tools are not well suited to manage large and complex amounts of sensitive data, as those collected in the cloud or exchanged through social networks. Ideally, we want to encrypt data while enabling fine-grained access control (for example, for selectively grant access to the data) and selective computation (for providing only partial access and performing operation directly over the encrypted data). That is, we want control over who as access to the encrypted data and what they can compute.
    In the recent years, functional encryption has emerged as a novel paradigm that addresses the above goals. I am interested in functional encryption from both the theoretical side (e.g. defining schemes for expressive functionalities and studying their security properties, for example) and from the experimental side (that is, implementing and deploying them in real-world settings, such as IoT or scada/ics scenarios and assessing their efficiency).
  • Involved person:  Alberto Trombetta
Data Privacy
  • Scrutinizing thoughts to design non-traditional solutions for a better understood, better managed, and better owned information privacy for all.  The goal is to deploy privacy-preserving solutions able to trade-off between privacy guarantees, utility of the managed data, and efficiency of data processing.  We are doing research on privacy-preserving techniques for a variety of application domains, including: data Streams, Online social networks, relational data management system, No SQL databases. Data Streams.
  • Involved person:  Barbara Carminati, Elena Ferrari
Deep Learning for Computer Vision and Natural Language Processing
  • Over the last few years Deep Learning is applied to various problems ranging from Computer Vision to Natural Language Processing and in many cases it outperformed previous state-of-the-art works. We are interested in developing new Deep Learning methods and applications for academic and real-world problems. We work on these business cases in various industries like online business, control systems, financial and educational. We are always interested in studying new fascinating problems and solving them with the use of Deep Learning.
  • Involved person: Ignazio Gallo
Empirical Software Engineering
  1. Software faultiness estimation: definition of practically useful discrete software faultiness estimation models built by using sensible thresholds for software measures on statistically significant continuous software fault-proneness models; use of machine-learning and conventional data analysis techniques; definition and assessment of prediction accuracy indicators.
  2. Software effort estimation: building of effort estimation models; evaluation of the accuracy of different models for the estimation of software development effort; definition and assessment of accuracy indicators.
  3. Functional size measurement: simplification and automation of measures for quantifying functional software size; investigation of their correlation with relevant software qualities, with special reference to development effort.
  4. Software code understandability: identification of factors that affect understandability, construction of understandability models and definition of new metrics to represent code properties (like complexity) that are expected to affect code understandability.
  • Involved persons:  Luigi Lavazza, Sandro Morasca
Formal Languages
  • Two main research areas. The former regards both enumerative combinatorics and combinatorial generation through efficient (Constant Amortized Time) algorithms based on discrete dynamical systems. The latter deals with formal languages and their analytic properties. With respect to the first area, we are currently developing CAT algorithms for the exhaustive generation of several classes of polyominoes and other combinatorial objects related to the so called Sand Pile Model. With respect to formal languages, our goal is that of looking for suitable classes of languages with the property of admitting generating functions belonging to particular classes of functions defined by means of linear differential equations with polynomial coefficients. This property is of particular interest since it allows to develop automatic tools for many classical decision problems.
  • Involved person: Paolo Massazza
Pattern Recognition and Computational Intelligence for Biomedical and Web Applications
  • Main research interests focus on Pattern Recognition and Computational Intelligence techniques with application to Biomedical Image Analysis, Medical Expert Systems and Web Data Mining.  The research activity is principally developed within the Center of Research in Image Analysis and Medical Informatics, with the main objective of promoting interdisciplinary research at the intersection of Computer Science and the Biomedical Field.  Ongoing projects focus on Tumour Segmentation of MRI Brain Scans using Energy Minimization and Machine Learning techniques, Clustering fMRI time series based on Fuzzy and Self-Organizing Neural methods for Brain Parcellation, Type1 and Type2 Fuzzy Logic Systems for Decision Support in Clinical Setting. Current research interests are also focusing on Conversational Analytics tools to improve human-to-bot interaction in Healthcare and Public Administration domains.
  • Involved person: Elisabetta Binaghi
  • The main research interests are focused on security and privacy solutions targeted to the Internet of Things (IoT) domain and Wireless Sensor Networks. Wireless sensor networks are characterized by poor resources which influence power consumption, congestion states, and possible loss of information. We proposed methods based on encryption techniques, hashing, and reputation schemas to cope with such issues. Moreover, we envisioned attribute-based access control techniques and policy enforcement mechanisms to guarantee security requirements in heterogeneous IoT scenarios. In fact, the diffusion of a huge amount of devices, adopting different network protocols and standards, emerges the need to introduce middleware layers, able to handle interoperability and scalability issues.  We also investigate intrusion detection techniques, based on Artificial Intelligence (AI), in order to promptly recognize malicious attacks and network misbehavior.
  • Involved persons: Alberto Coen Porisini, Alessandra Rizzardi, Sabrina Sicari
Software metrics and software quality evaluation
  • Software project management and effort estimation; Software process modeling, measurement and improvement. Definition and evaluation of estimation models; to this end, techniques for setting thresholds (e.g., to tell apart faulty and not faulty software modules) are investigated. Software quality front, issues concerning code smells and technical debt identification, measurement and management. Open Source Software; Requirements engineering; Model based development, especially concerning real-time and embedded software.
  • Involved person: Luigi Lavazza
Trust & Risk
  • We are interested in trust as a measure of the confidence that an entity will behave in an expected manner with respect to the security/privacy policies in place.  In addition to trust measures, we are interested in studying  risk that can be seen as the other side of trust.  At this aim,  we are interested to study  the concept of risk one might be exposed to when interacting with other  users in an web environment. The above considerations lead us to investigate  models and estimation methods for trust and risk  in a variety of application domains, including: mobile payment, trust&risk in online social networks, trust in decentralized social networks.
  • Involved persons:  Barbara Carminati, Elena Ferrari