PhD in Computer Science

Discussioni Tesi di Dottorato Marengo e Pomponiu

Academic year 2012/2013

Sommario del corso

• Course objectives
• Notes
• Teaching tools

Course objectives

ore 12-13: Elisa Marengo

Titolo: 2CL Protocols: Interaction Patterns Specification in Commitment Protocols.

Abstract: Commitment protocols have been deeply investigated in the context

    of multiagent systems (MAS) as a valuable way for specifying interaction 

    and communication protocols. Commitments capture social relations among 

    the interacting parties, expressing the engagement of a debtor towards a 

    creditor agent, to do something or to achieve some condition of interest. 

    By specifying only what agents are expected to do, rather then how they 

    are expected to satisfy their commitments, this kind of specification is highly 

    appreciated because it leaves great freedom of behaviour to agents. 

    However, in many practical contexts there is the need of capturing some 

    patters that the interaction is desired to respect and that improve coordination. 

    The contribution of this thesis goes in this direction, by presenting an extension 

    of commitment protocols which allows one to express a set of desired patterns

    given as constraints among facts and commitments. Being the approach

    declarative, it still provide a high flexibility to agents. More in detail, the proposal 

    is to explicitly account for a regulative component of the specification, capturing 

    the set of constraints on the interaction, and a separate and decoupled constitutive 

    specification defining the set of actions that can be performed. The advantages 

    of this decoupling is a greater modularity in the specification which entails an 

    easier protocol design, re-use and extension. An operational semantics is also 

    provided. It is obtained as extension of the commitment machine 

    in such a way to consider the role of constraints in deciding which of the possible

    interactions can be considered as legal for the protocol.

    The proposal is validated by modelling some real case studies and by

    showing the advantages of its adoption with respect to other proposals from 

    the literature. In order to support the design and the analysis of protocol 

    specifications, a tool offering different kinds of functionalities is presented. 

    Among these, a prolog program allows for the generation of the labelled 

    graph of the possible interactions, where labels are used to represent constraints 

    or commitments violations.

ore 13-14: Victor Pomponiu

Titolo: The Use of Singular Value Decomposition in Information Hiding

Abstract: In recent years, the necessity to protect multimedia content from 

    illegal copying has been made more critical by the advent of digital technology. 

    A common and well-discussed solution to counter the unauthorized distribution 

    of copyrighted contents is applied by means of digital watermarking. This term 

    refers to specific information hiding techniques whose purpose is to embed secret 

    information inside multimedia contents, such as images, video or audio streams. 

    The watermark, i.e., the signal added to digital media, can be detected and retrieved 

    when necessary. In the specific field of content protection, the objective is to identify 

    the media’s owner by means of a specific user-related watermark.

     In the last years the well-known numerical tool called Singular Value Decomposition (SVD) 

    received much attention from the watermarking community. However, few of these 

    schemes take into consideration another crucial constraint, the security, which 

    guarantees the usability of the watermarking technology. This dissertation focuses on 

    the security of digital watermarking applications that make use of SVD transform to 

    convey the watermark information.

     First, we analyze and quantify the security of the watermarking schemes based on SVD 

    that embed the watermark additively of through quantization. These schemes are insecure 

    and cannot be used for applications that require ownership protection and authentication. 

    This is due to the fact that the space where the watermark resides, i.e., the embedding 

    space, can be estimated and the watermark removed or overwritten. To motivate our 

    findings we devise two novel attacks that remove the watermark while minimizing the 

    perceptual distortion.

    To cope with these attacks we explore the use of the singular vectors for information 

    hiding. In particular, the proposed watermarking algorithm, that outperforms previously 

    proposed image watermarking methods, inserts the watermark by modifying the angles 

    formed by the singular vectors. 

Class schedule

Note

La commissione sarà formata da:

Prof.ssa Carla Simone (Università degli Studi di Milano-Bicocca) 

Prof. Enrico Magli (Politecnico di Torino) 

Dr. Paolo Torroni (Università di Bologna) 

• Teaching materials
• Exams and finals
• Go to Moodle
• Visit the forums