Ran Yahalom

Structure-based identification of catalytic residues†

The identification of catalytic residues is an essential step in functional characterization of enzymes. We present a purely structural approach to this problem, which is motivated by the difficulty of evolution‐based methods to annotate structural genomics targets that have few or no homologs in the databases. Our approach combines a state‐of‐the‐art support vector machine (SVM) classifier with novel structural features that augment structural clues by spatial averaging and Z scoring. Special attention is paid to the class imbalance problem that stems from the overwhelming number of non‐catalytic residues in enzymes compared to catalytic residues. This problem is tackled by: (1) optimizing the classifier to maximize a performance criterion that considers both Type I and Type II errors in the classification of catalytic and non‐catalytic residues; (2) under‐sampling non‐catalytic residues before SVM training; and (3) during SVM training, penalizing errors in learning catalytic residues more than errors in learning non‐catalytic residues. Tested on four enzyme datasets, one specifically designed by us to mimic the structural genomics scenario and three previously evaluated datasets, our structure‐based classifier is never inferior to similar structure‐based classifiers and comparable to classifiers that use both structural and evolutionary features. In addition to the evaluation of the performance of catalytic residue identification, we also present detailed case studies on three proteins. This analysis suggests that many false positive predictions may correspond to binding sites and other functional residues. A web server that implements the method, our own‐designed database, and the source code of the programs are publicly available at http://www.cs.bgu.ac.il/∼meshi/functionPrediction. Proteins 2011;

Mining roles from web application usage patterns

Role mining refers to the problem of discovering an optimal set of roles from existing user permissions. In most role mining algorithms, the full set of user-permission assignments (UPA) is given as input. The challenge we are facing in the current paper is mining roles from actual web-application usage information. This information is collected by monitoring the access of users to application during a period of time. We analyze the actual permissions required to access the application in each users session, and construct a set of user-permission assignments, which result in an incomplete UPA. We propose an algorithm that uses the session permission information to overcome the deficient data. We show by example how each step of the algorithm overcomes by heuristic instances of higher uncertainty. We demonstrate by simulation the efficiency of our algorithm in handling different levels of deficient data.

Constrained anonymization of production data: a constraint satisfaction problem approach

The use of production data which contains sensitive information in application testing requires that the production data be anonymized first. The task of anonymizing production data becomes difficult since it usually consists of constraints which must also be satisfied in the anonymized data. We propose a novel approach to anonymize constrained production data based on the concept of constraint satisfaction problems. Due to the generality of the constraint satisfaction framework, our approach can support a wide variety of mandatory integrity constraints as well as constraints which ensure the similarity of the anonymized data to the production data. Our approach decomposes the constrained anonymization problem into independent sub-problems which can be represented and solved as constraint satisfaction problems (CSPs). Since production databases may contain many records that are associated by vertical constraints, the resulting CSPs may become very large. Such CSPs are further decomposed into dependant sub-problems that are solved iteratively by applying local modifications to the production data. Simulations on synthetic production databases demonstrate the feasibility of our method.

CAMLS: a constraint-based apriori algorithm for mining long sequences

Mining sequential patterns is a key objective in the field of data mining due to its wide range of applications. Given a database of sequences, the challenge is to identify patterns which appear frequently in different sequences. Well known algorithms have proved to be efficient, however these algorithms do not perform well when mining databases that have long frequent sequences. We present CAMLS, Constraint-based Apriori Mining of Long Sequences, an efficient algorithm for mining long sequential patterns under constraints. CAMLS is based on the apriori property and consists of two phases, event-wise and sequence-wise, which employ an iterative process of candidate-generation followed by frequency-testing. The separation into these two phases allows us to: (i) introduce a novel candidate pruning strategy that increases the efficiency of the mining process and (ii) easily incorporate considerations of intra-event and inter-event constraints. Experiments on both synthetic and real datasets show that CAMLS outperforms previous algorithms when mining long sequences.

USB-based attacks

Abstract Attackers increasingly take advantage of innocent users who tend to use USB peripherals casually, assuming these peripherals are benign when in fact they may carry an embedded malicious payload that can be used to launch attacks. In recent years, USB peripherals have become an attractive tool for launching cyber-attacks. In this survey, we review 29 different USB-based attacks and utilize our new taxonomy to classify them into four major categories. These attacks target both individuals and organizations; utilize widely used USB peripherals, such as keyboards, mice, flash drives, smartphones etc. For each attack, we address the objective it achieves and identify the associated and vulnerable USB peripherals and hardware.

Identifying Knots of Trust in Virtual Communities

Knots of trust are groups of community members having overall “strong” trust relations between them. In previous work we introduced the knot aware trust based reputation model. According to this model, in order to provide a member with reputation information relative to her viewpoint, the system must identify the knot to which that member belongs and interpret its reputation data correctly. In the current paper we present the problem of identifying knots which is modeled as a graph clustering problem, where vertices correspond to individuals and edges describe trust relationships between them. We propose a new perspective for clustering that reflects the subjective idea of trust and the nature of the community. A class of weight functions is suggested for assigning edge weights and their impact on the stability and strength of knots is demonstrated. Finally we show the efficiency of knots of high quality for providing their members with relevant reputation information.

Constrained obfuscation of relational databases

The need to share data often conflicts with privacy preservation. Data obfuscation attempts to overcome this conflict by modifying the original data while optimizing both privacy and utility measures. In this paper we introduce the concept of Constrained Obfuscation Problems (COPs) which formulate the task of obfuscating data stored in relational databases. The main idea behind COPs is that many obfuscation scenarios can be modeled as a data generation process which is constrained by a predefined set of rules. We demonstrate the flexibility of the COP definition by modeling several different obfuscation scenarios: Production Data Obfuscation for Application Testing (PDOAT), anonymization of relational data, and anonymization of social networks. We then suggest a general approach for solving COPs by reducing them into a set of Constrained Satisfaction Problems (CSPs). Such reduction enables the employment of the well-studied CSP framework in order to solve a wide range of complex rules. Some of the resulting CSPs may contain a large number of variables, which may make them intractable. In order to overcome such intractability issues, we present two useful heuristics that decompose such large CSPs into smaller tractable sub-CSPs. We also show how the well-known @?-diversity privacy measure can be incorporated into the COP framework in order to evaluate the privacy level of COP solutions. Finally, we evaluate the new method in terms of privacy, utility and execution time.