ANONITECH LLC — Department of Energy STTR Phase I: 05a

Description

Modern network devices collect a huge amount of network data that can be analyzed to understand how the system performs, identify bottlenecks and anomalies, detect attacks, and analyze network hierarchy. Therefore, there is a need for analysis of network data either in-house, or by an external organization that specializes in such analyses, or by the research community. However, for the network data to be shared, it must first be anonymized due to privacy concerns. Few existing anonymization methods allow enough privacy protection (organization need), or acceptable data utility (researcher need), or efficient data analysis but no method achieves all, at the same time. This project will develop an innovative privacy management tool to anonymize packet header data, network flow data, or TSTAT data and achieve all these goals. The core of the proposed solution is a novel condensation-based differential privacy method. The proposed solution is innovative in the following aspects: 1) a strong privacy guarantee based on differential privacy that is immune to various attacks against anonymization including injection attacks; 2) better preservation of data properties that enables subsequent analysis compared to existing anonymization techniques due to an innovative condensation-based differential privacy clustering method as well as an intelligent tuning mechanism that makes this tool adaptive to different domain-specific contexts including the re-identification risks of a field in the packet header and its importance in analysis tasks; 3) no extra computation overhead for the data analyzer, i.e., the data analyzer can conduct data analysis directly on the anonymized data without the need of decryption. Specifically, during phase I, the project team will perform a proof-of-principle demonstration for a system with the following deliverables: 1) a novel condensation-based differential privacy anonymization tool that can anonymize GBs of network header data in minutes; 2) an automatic tuning mechanism that can capture user context and optimize preservation of data properties important for analysis as well as providing adequate privacy protection; 3) extensive comparison with existing anonymization tools/methods on various types of network data and analysis tasks showing significant improvement by our methods over existing ones on both privacy protection and preservation of data properties useful for subsequent analysis; 4) an integrated prototype with an intuitive user interface that is easy to use and provides quick insight of both original and anonymized data. In phase II and beyond, the team will concentrate on a widespread adoption of the proposed anonymization technology by as many organizations as possible. It will make organizations feel more open to distributing their data for analysis without any fear of divulging sensitive information. We plan to expand to additional datasets beyond network, such as healthcare data benefiting directly the organizations and indirectly the society on private and safe data usage via anonymization. The global data anonymization market size is expected to grow from USD 384.8 million in 2017 to USD 767.0 million by 2022, at a Compound Annual Growth Rate (CAGR) of 14.8%. The proposed solution will lead to a product that addresses the risk in exporting data which benefits both the data owners (business or government agency) and the data recipients (contractors, licensees, or researchers). For data recipients, the tool preserves the utility of the data but also mitigates the risk of not meeting the terms and conditions of the data usage contract, a problem several companies have experienced. A breach of privacy is difficult to recover from, no matter how the breach occurs, and the loss of trust can destroy a business or its reputation.