Entity

Time filter

Source Type

Columbia, MD, United States

Das K.,NASA | Bhaduri K.,Critical Technologies Inc | Kargupta H.,University of Maryland Baltimore County | Kargupta H.,Agnik LLC
Knowledge and Information Systems | Year: 2010

In this paper we develop a local distributed privacy preserving algorithm for feature selection in a large peer-to-peer environment. Feature selection is often used in machine learning for data compaction and efficient learning by eliminating the curse of dimensionality. There exist many solutions for feature selection when the data are located at a central location. However, it becomes extremely challenging to perform the same when the data are distributed across a large number of peers or machines. Centralizing the entire dataset or portions of it can be very costly and impractical because of the large number of data sources, the asynchronous nature of the peer-to-peer networks, dynamic nature of the data/network, and privacy concerns. The solution proposed in this paper allows us to perform feature selection in an asynchronous fashion with a low communication overhead where each peer can specify its own privacy constraints. The algorithm works based on local interactions among participating nodes. We present results on real-world dataset in order to test the performance of the proposed algorithm. © 2009 Springer-Verlag London Limited. Source


Grant
Agency: Department of Defense | Branch: Army | Program: STTR | Phase: Phase I | Award Amount: 100.00K | Year: 2007

This proposal suggests research on security issues for mobile databases running on a mobile ad hoc network (MANET). It will explore both state of the art commercial systems and advanced research issues for mobile security management. The first step will involve an extensive study of the commercial off-the-shelf technology. This study will identify the resources that can be used for addressing security issues in a MANET. This research will also explore several other issues that are important for this problem. Examples include data relocation/distribution and storage strategies for data confidentiality and fault tolerance, local agreement detection, secure update propagation, collaborative query processing, situation awareness and profile-based monitoring. The proposed research will explore several advanced data analysis-based techniques for network traffic monitoring and outlier detection. This work will be jointly performed by Agnik, LLC and University of Maryland, Baltimore County.


Grant
Agency: Department of Defense | Branch: Missile Defense Agency | Program: SBIR | Phase: Phase II | Award Amount: 999.99K | Year: 2009

This work develops Distributed Information Assurance (DIA) system based on the distributed data mining technology. The key capabilities are: ++Multi-agent architecture linking multiple, heterogeneous network-sensors to perform distributed and decentralized analysis of the data, supporting local management of the policy-based control for different sensors. ++Collection of distributed data mining algorithms. ++Module for supporting the complete life-cycle of the information assurance management process in a BMDS including information discovery, linking cross-domain information, policy management, and effective utilization of the knowledge. ++Detect (a) distributed network “signatures” of attackers based on distributed observations from different nodes in the network. (b) attack patterns on different components of the BMDS network in terms of clusters, outliers and also identify statistical properties of attack distribution in order to perform a trend analysis. (c) stealth network probes by attackers and worms (d) insider attacks on the BMDS network. The DIA system will be interfaced with existing third-party network sensors such as network intrusion detection systems (e.g. SNORT), host-based intrusion detection systems (OSSEC), router logs (e.g. CISCO Netflows), network and personal firewalls. ++Web-service based service-oriented architecture for quick intervention from the administrators, distributed collaboration among peers supporting analysis of a threat by allowing case archival and case-based reasoning.


Grant
Agency: Department of Defense | Branch: Missile Defense Agency | Program: SBIR | Phase: Phase I | Award Amount: 99.96K | Year: 2008

This document proposes to develop a Distributed Information Assurance (DIA) system based on the distributed data mining technology for detecting distributed network attacks and identifying attackers’ “signatures” for advanced situational awareness. It will offer the following key capabilities: 1. A multi-agent architecture for linking multiple, heterogeneous network-sensors (e.g., intrusion detection and malware detection systems, netflow data, tcpdump) for performing distributed and decentralized analysis of the data. The system will support local management of policy-based control for different sensors. 2. A collection of distributed data mining algorithms for decentralized outlier detection, clustering,and trend analysis for network data. These algorithms will lay the foundation of the DIA system. 3. A module for supporting the complete life-cycle of the information assurance management process in a BMDS. Following attack detection capabilities will be explored during Phase I: Detect distributed network “signatures” of attackers based on the distributed observations from different nodes in the network. Detect attack patterns on the coalition members in terms of clusters, outliers. Identify statistical properties of attack distribution in order to perform a trend analysis. Detect stealth network probes by attackers and worms. The proposed work will be performed at Agnik, a mobile and distributed data mining company.


Branch J.W.,IBM | Giannella C.,Mitre Corporation | Szymanski B.,Rensselaer Polytechnic Institute | Wolff R.,Haifa University | And 2 more authors.
Knowledge and Information Systems | Year: 2013

To address the problem of unsupervised outlier detection in wireless sensor networks, we develop an approach that (1) is flexible with respect to the outlier definition, (2) computes the result in-network to reduce both bandwidth and energy consumption, (3) uses only single-hop communication, thus permitting very simple node failure detection and message reliability assurance mechanisms (e. g., carrier-sense), and (4) seamlessly accommodates dynamic updates to data. We examine performance by simulation, using real sensor data streams. Our results demonstrate that our approach is accurate and imposes reasonable communication and power consumption demands. © 2012 Springer-Verlag London Limited. Source

Discover hidden collaborations