Maurizio Panti

A conceptual indexing method for content-based retrieval

A great deal of work has been done to define index structures to support feature-based similarity queries. However, other kinds of content-based retrieval, namely keyword-based and concept-based, are founded on different properties of the data space, which make these methods ineffective. Nevertheless, similarity notions are still needed, in order to manage incompleteness and imprecision in the representation of multimedia data, as well as in user query specification. In the paper, we present an indexing method which is based on partitioning the data space. We introduce the binary counterpart of the notions of minimum volume and minimum overlap, and combine them in a global hierarchical clustering criterion. We also show how the index structure induced by the clusterization can be exploited to deal with incompleteness and imprecision expressed in terms of answer precision and recall.

Cooperation Strategies for Information Integration

We discuss and analyse cooperation strategies for rewriting queries in a mediator architecture. According to this approach, the mediated schema of each mediator is dynamically updated through the cooperation with information sources and other mediators, strongly influenced by the queries submitted by a consumer. This approach allows a mediator to face systems where information sources and consumer needs are dynamic. From the analysis of different cooperation strategies arises that it is more efficient and effective to directly cooperate with information sources when the sources are few. Otherwise, it is more efficient to cooperate with other mediators.

A Logic of Belief and a Model Checking Algorithm for Security Protocols

Model checking is a very successful technique which has been applied in the design and verification of finite state concurrent reactive processes. In this paper we show how this technique can be used for the verification of security protocols using a logic of belief. The underlying idea is to treat separately the temporal evolution and the belief aspects of principals. In practice, things work as follows: when we consider the temporal evolution of a principal we treat belief atoms (namely, atomic formulae expressing belief) as atomic propositions. When we deal with the beliefs of a principal A, we model its beliefs about another principal B as the fact that A has access to a representation of B as a process. Then, any time it needs to verify the truth value of some belief atom about B, e.g., B B o, A simply tests whether, e.g., o holds in its (appropriate) representation of B. Beliefs are essentially used to control the “jumping” among processes. Our approach allows us to reuse the technology and tools developed in model checking.

Developing an open knowledge discovery support system for a network environment

Knowledge discovery in databases (KDD) is a highly complex process where a lot of data manipulation tools with different characteristics can, and in fact have to, be used together in an interactive and iterative fashion, to reach the goal of previously unknown, potentially useful information extraction. In this paper we analyze the major sources of complexity in the framework of network organizations, pointing out the necessity to give support to the user in many different ways and at very different levels of granularity, from the use of a single tool, to the management of whole, distributed, KDD projects. Unfortunately, currently available systems lack to support the users in at least some of these features. We then propose a solution based on the service oriented computing paradigm, arguing that the advantages of this paradigm, namely openness, modularity, reusability and transparency, as well as ubiquity, can help in the design of an effective support system for knowledge discovery in databases in network environments

Agent-based transactions into decentralised P2P

The Peer-to-Peer (P2P) approach is considered an innovative paradigm in distributed computing. Nevertheless, P2P tools are not able to exchange complex data and to deal with heterogeneity and data management problems. In order to overcome some of these limitations, this paper proposes to use an agent based cooperative information system. We describe the fundamental agent roles inside the system: mediator, facilitator, planner and wrapper. In particular, the mediator plays the main role in generating simple data management operations (transactions) according to its local knowledge.

An efficient and scalable data compression approach to classification

Learning algorithms are effective means of inducing predictive models of a phenomenon starting from a set of instances of the phenomenon itself. However, the impressive growth of the amount of stored data makes scalability of both the learning and classification procedures a compelling requisite for their effective application in data mining tasks, at least as important as accuracy of the induced model. In this paper we show the features of a stochastic gradient algor i thm for the minimization of the average misclassification risk performed by a Labeled Vector Quantizer, both in terms of scalability and accuracy. The performance are compared with those of other related algorithms, often adopted in data mining, on both artificial and real data experiments.

Model checking the security of multi-protocol systems

Many security protocols have been shown to have serious design flaws, even a long time after their publication. For this reason, several automatic verification methods have been proposed. As far as we know, all these methods deal with the verification of mono-protocol systems. On the contrary, modern computer systems employ different security protocols in the same environment. In this paper, we present a methodology to automatically verify the security of multiprotocol systems based on model checking. We also describe its application to a multi-protocol system composed by SET (by VISA & Mastercard) and 2KP (by IBM)

Verification of the SSL/TLS Protocol Using a Model Checkable Logic of Belief and Time

The paper shows how a model checkable logic of belief and time (MATL) can be used to check properties of security protocols employed in computer networks. In MATL, entities participating to protocols are modeled as concurrent processes able to have beliefs about other entities. The approach is applied to the verification of TLS, the Internet Standard Protocol that IETF derived from the SSL 3.0 of Netscape. The results of our analysis show that the protocol satisfies all the security requirements for which it was designed.

Modeling and evaluating cooperation strategies in P2P agent systems

Distributed computing is becoming a fundamental technology for information exchange and cooperation. However, for such a technology to gain wide use, it must cater to users who do not own sophisticated (hardware and software) platforms and permanent network connections. For such users, the recent Peer-to-Peer (P2P) computing model has many advantages over classical client-server and web-based distributed models. However, the P2P computing model in its current form has a number of other limitations in the data exchange and the protocols it supports. To overcome some of these, we proposed an agent based P2P system model whose nodes are agents (Peer Agents). In this paper, we adopt the i* graphical framework to help the modeling and evaluating of Peer Agents cooperation strategies. Although, the highly dynamic nature of P2P networks complicates the evaluation procedure, we propose three possible evaluation criteria in order to characterize the best strategy related to a particular failure symptom.

A modulus oriented hash function for the construction of minimal perfect tables

Purpose of this paper is to describe an algorithm for the construction of minimal perfect hash tables for small static word sets. As first step we have analysed Cook Oldehoefts letter oriented algorithm and then we have introduced some modifications to reduce the execution time and to improve the success per tries ratio. In the first two paragraphs we briefly describe the basic algorithm and then the modifications adopted. Successively, we describe the data and the parameters defined to evaluate the performances of the algorithms. The last paragraph contains the discussion of the results obtained.