SCENS: The Secure Content Exchange Negotiation System

SCENS, the Secure Content Exchange Negotiation System, is an NSF-funded data access and sharing system based on the idea of "mediating" exchanges between parties using a central server. A key element of the SCENS approach is the introduction of a paradigm of "mediated sharing", in which participants share only metadata descriptions of their original data with a trusted central server (SCENS). The SCENS server then facilitates matching interested parties, assists parties in negotiating on access conditions, records exchange agreements, and collects and monitors feedback from participants.

This website includes the objectives, challenges, result and deliverable (development and evaluation), highlights, and publications of the SCENS project. SCENS is a part of the NSF project, ITR: A System for Data Integration and Pattern Discovery in Multimodal, Spatio-temporal Data: Lesion Analysis and Data Sharing, but focusing on data sharing mechanisms.

Acknowledgement: This material is based in part upon work supported by the National Science Foundation under Grant Numbers IIS-0733674 (the original award IIS-0312629). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.

PI, co-PI(s): Makedon, Fillia (PI)
                   Pearlman, Justin (Co-PI)

                   Ford, James (Co-PI)
                   Saykin, Andrew (Co-PI)
                   Wishart, Heather(Co-PI)

Senior Personnel(s):
                   Simpkins, James
                   Kristen, Chambers
                   Bishop,Matt
                  
Student(s):
                   Li Shen (now an assistant professor in Indiana University)
                   Yuhang Wang (now an assistant professor in Southern Methodist University)
                   Tilmann Steinberg (Graduated)
                   Heng Huang (now an assistant professor in UT Arlington)
                   Ye Song (Graduated)
                   Zhengyi Le (Now in UT Arlington)                  
                   Yi Ouyang (Graduated)
                   Zhifeng Wang (Graduated)                  
                   Yurong Xu (now in UT Arlington)
                   Sheng Zhang (Graduated)
                   Wei Zheng (Graduated)
                   Rong Zhang (Graduated)
                   Eric W. Becker (now in UT Arlington)                 
                   Vangelis Metsis (now in UT Arlington)

Objectives

There were three objectives that drove the development of SCENS:

  1. System Development: Develop a prototype data management system to manage the collection, searching, evaluation and negotiation of shared data within and across research entities or institutions.
  2. System Evaluation: Evaluate system usefulness in the context of brain imaging and mental health data (using synthetic data).
  3. Research: Research security issues related to the proposed data management and negotiation framework.

Research Challenges

When a data owner likes to share his/her data sets though Internet, the major concerns are, (1) how can I take benefits from sharing my data?, (2) how secure it is? and (3) will it be very slow? So, the challenges that data sharing research is facing are exchange mechanism, security and efficiency.

Result and Deliverable

SCENS System Development The prototype uses a Java-based Web interface, supports interactive and offline negotiations between human users, and provides collaborative filtering tools for information access.

In parallel, SCENS has been extended to support autonomous clients in a P2P framework using trust negotiation, a method for building trust between parties through exchanges of credentials. We proposed a new variant of this approach called "collaborative trust negotiation" that allows more than two parties to participate in order to improve the likelihood of completing negotiations. Generalization from a client-server to a P2P model eliminates a key potential bottleneck in the SCENS approach, namely the use of a central server, and potentially can provide a significant boost to efficiency (dependent on the negotiation patterns of the participants). For more information on this P2P extension, see our [IEEE P2P 2004] paper.

Publications: [ACM EC 2004, EC-WEB 2004, SETN 2004, IEEE P2P 2004, IEEE CEC 2005, SDM2006a, PCI 2007]

SCENS Evaluation We have evaluated the system using synthetic medical imaging data and simulations; in addition, we have introduced a second problem domain, sharing computer code, in order to introduce issues related to intellectual property. We are now also preparing to use a third domain using user-tagged image files.

Use of the prototype system has allowed us to collect negotiation history data for our negotiation strategy research and user ratings and logs for our work on recommendation systems and trust building. This also stimulated several derivative lines of research related to improving the efficiency, security, and privacy of the collaborative filtering tools used to generate recommendations for SCENS users.

Publications: [IEEE P2P 2004, ACM MM 2004, Zhao MS 2005]

SCENS Research We have reported results on multiple aspects of data security; the three highlights are a new approach to multi-party access control based on digital credentials, new detection and defense mechanisms related to attacks on the mechanism for recommending data, and new approaches for preserving privacy in collaborative filtering.

Publications: [EGOV 2004, ACM SAC 2005, ACM EC 2006, SDM 2006b, ACM SIGIR 2006, ACM KDD 2006]

Broader Impacts

The expansion of this project to include data gathering from the home through sensor-based assisted living deployments is likely to be an excellent topic for future commercialization, as the same mechanisms designed for gathering research and clinical data can provide support to caregivers.

Figure 1: The three layers (service types) offered by SCENS.

Figure 2: SCENS components.

SCENS Highlights

We have developed an incentive mechanism for the P2P version of SCENS to urge selfish peers to collaboratively share their bandwidth in P2P media streaming systems. It combines the traditional reputation-based approach and an online streaming behavior monitoring scheme. The incentive mechanism used by this method is orthogonal to existing media streaming solutions and can be integrated into them. Our preliminary results show that when our bandwidth sharing system is applied, the overall performance achieved by collaborative peers in our P2P media streaming application does not suffer from the existence of non-collaborative peers [ACM MM 2004]. There is a potential to apply SCENS to other similar problems, including sharing of computational resources, such as CPU cycles and disk space.

We have designed efficient collaborative filtering (CF) techniques to help users find interesting information and items rapidly from a large amount of candidates in SCENS and other data sharing systems [PCI 2007]. We have developed an approach based on using Singular Value Decomposition (SVD) approximation in an SVD-based CF algorithm to reduce the computational cost [IEEE CEC 2005] and an approach based on using non-negative matrix decomposition to efficiently categorize user interests [SDM 2006a].

We have successfully used our visualization tool to identify parties with different negotiation behavior characteristics in simulated negotiation transactions. Our preliminary results indicate that it can be useful for supporting detection of two important changes to the global negotiation environment: changes in the popularity of a dataset and changes in the behavior of a user. The tool can also direct users to the parties that they are calculated to have the best chance of successful negotiation with in the future based on global statistics. This work formed a substantial part of a Computer Science Master's thesis [Zhao MS 2005].

We have designed a negotiation strategy pool framework that allows automated negotiation agents to select strategies from a range of pre-evaluated options [ACM EC 2004]. This approach allows for more sophisticated negotiation on the part of negotiation agents. We have also designed a negotiation protocol for multi-party negotiations based on majority rule [EC-WEB 2004], and have done preliminary work on creating new techniques for negotiation agents to use in securely sharing their past negotiation data to support machine learning of negotiation strategies [EGOV 2004]. The main difficulty in this work is preserving the privacy of internally held information on negotiations.

We have designed a privacy-preserving negotiation learning framework to allow negotiation parties to employ learning techniques on the union of their past negotiation records without compromising their privacy [ACM SAC 2005]. We also published a novel distributed scheme in which users maintain their own rating profiles for privacy and the server periodically collects aggregate information from online users to provide recommendations [IEEE CEC 2005] and an extension in which the server augments this approach by providing 'hints' to clients [ACM EC 2006]. In conjunction with that work, we determined how vulnerable various approaches to protection of privacy based on random perturbations are [SDM 2006b] and how standard attack models on CF systems worked against both existing approaches and our new methods [ACM SIGIR 2006]. Finally, we also devised a new attack detection approach based on evaluating changes to the system as time series [ACM KDD 2006].

Online Demo

To access an online demo of SCENS, click the following link:

http://heracleia.uta.edu:8080/SCENS/

You can use the account name 'guest' and password 'guest', or create a new account.

Publications

[ACM EC 2004] S. Zhang, S. Ye, F. Makedon, and J. Ford, "A Hybrid Negotiation Strategy Mechanism in an Automated Negotiation System" (extended abstract), ACM E-Commerce'04, New York City, May 17-20, 2004.

[ACM MM 2004] S. Ye and F. Makedon. "Collaboration-Aware Peer-to-Peer Media Streaming", The 12th International ACM Conference on Multimedia (MM 2004), pp. 41

[EC-WEB 2004] S. Zhang, F. Makedon, J. Ford, and L. Ai, "A Model for Multi-party Negotiations with Majority Rule", Fifth International Conference on E-Commerce and Web Technologies (EC-Web 2004), Lecture Notes in Computer Science 3182, Springer, pp. 228-237, Zaragoza, Spain, Aug. 30-Sep. 3, 2004.

[EGOV 2004] S. Zhang, F. Makedon, J. Ford, C. Sudborough, L. Ai, S. Kapidakis, V. Karkaletsis, and E. Loukis, "An International Trade Negotiation Framework for E-Government", Third International Conference on Electronic Government (EGOV '04), Lecture Notes in Computer Science 3183, Springer, pp. 211-217, Zaragoza, Spain, Aug. 30-Sep. 3, 2004.

[IEEE P2P 2004] S. Ye, F. Makedon, J. Ford, and L. Ai, "Collaborative Automated Trust Negotiation in Peer-to-Peer Systems", The Fourth IEEE International Conference on Peer-to-Peer Computing (IEEE P2P 2004), Zurich, Switzerland, August 25-27, 2004.

[SETN 2004] F. Makedon, S. Ye, S. Zhang, J. Ford, L. Shen, and S. Kapidakis, "Data Brokers: Building Collections through Automated Negotiation", Methods and Applications of Artificial Intelligence: the Third Hellenic Conference on Artificial Intelligence (SETN 2004), Lecture Notes in Artificial Intelligence 3025, Springer, pp. 13-24, Samos, Greece, May 5-8, 2004.

[Zhao MS 2005] Y. Zhao, SCENS: Supporting and Visualizing Negotiation Communications, Master's thesis, Dartmouth College, January 2005.

[ACM SAC 2005] W. Zheng, S. Zhang, Y. Ouyang, J. Ford and F. Makedon, "Node Clustering Based on Link Delay in P2P Networks", Proceedings of the 20th ACM Symposium of Applied Computing (SAC 2005), Santa Fe, New Mexico, March 14-17, 2005.

[IEEE CEC 2005] S. Zhang, W. Wang, J. Ford, and F. Makedon, "Using Singular Value Decomposition Approximation for Collaborative Filtering", The 7th IEEE Conference on E-commerce Technology (IEEE CEC 2005), Munchen, Germany, July 19-22, 2005.

[ACM EC 2006] S. Zhang, J. Ford, and F. Makedon, "A Privacy-Preserving Collaborative Filtering Scheme with Two-way Communication," ACM Conference on Electronic Commerce (EC06), Ann Arbor, MI, June 11-15, 2006.

[ACM KDD 2006] S. Zhang, A. Chakrabarti, J. Ford, and F. Makedon, "Attack Detection in Time Series for Recommendation Systems", The Twelfth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD 2006), Philadelphia, PA, August 20-23, 2006.

[ACM SIGIR 2006] S. Zhang, Y. Ouyang, J. Ford, and F. Makedon, "Analysis of a Low-dimensional Linear Model under Recommendation Attacks", The 29th Annual International ACM Conference on Research & Development on Information Retrieval (SIGIR 2006), Seattle, WA, August 6-11, 2006.

[SDM2006a] S. Zhang, J. Ford, and F. Makedon, "Learning from Incomplete Ratings Using Non-negative Matrix Factorization", The 2006 SIAM Conference on Data Mining (SDM06), Bethesda, MD, April 20-22, 2006.

[SDM 2006b] S. Zhang, J. Ford, and F. Makedon, "Deriving Private Information from Randomly Perturbed Ratings", The 2006 SIAM Conference on Data Mining (SDM06), Bethesda, MD, April 20-22, 2006.

[PCI 2007] F. Makedon, S. Zhang, Z. Le, J. Ford, and E. Loukis, "Providing Recommendations in an Open Collaboration System", The 11th Panhellenic Conference on Informatics (PCI 2007), Patras, Greece, May 18-20, 2007. To appear.

Date of Last Update: 2008.7.28
For SCENS source code, please contact: Rong Zhang or Song Ye