zur Startseite

Masterarbeit

Speculative Reordering for a Latency-optimized Privacy Protection in Complex Event Processing
Betreuer M. Sc. Saravana Murthy Palanisamy
Prüfer Prof. Dr. rer. nat. Dr. h. c. Kurt Rothermel
Beschreibung

1 Background
Complex Event Processing (CEP) is a state-of-the-art paradigm in the world of IoT to transform streams of raw data collected from sensors to meaningful information that are required to offer IoT services. But, data collected from these sensors often contain sensitive information about the user and are hence privacy-sensitive. Thus for IoT services to be widely accepted, protecting privacy is of paramount importance. According to a YouGov survey, 32% of the participants would be willing to measure and share fitness data with their health insurance provider[1] e.g., to get a premium rebate. However, 73% are afraid that their premiums would increase, since the  insurers could infer a certain lifestyle or disease from the shared data. Access control is one prominent technique for such privacy protection, but most access control mechanisms protect privacy only at the level of single attributes of data or events. However, sensitive information is often revealed via patterns of events. We call these privacy-sensitive event patterns as private patterns. In contrast, public patterns are non-privacy sensitive and are necessary to offer various IoT services. Thus private patterns should be concealed while preserving maximum number of public patterns to achieve a better Quality of Service (QoS). We already developed such a pattern-level access control component based on reordering of events. This component performs reordering window-by-window. Also, we reorder only after all the events in the current time window are available since it is necessary to know all public and private pattern matches before reordering.


2 Research Question and Goals
Although reordering is simple and fast, the processing time for reordering increases significantly when the window size increases. Therefore, in this thesis, we intend to develop a speculative strategy that predicts the completion of private patterns and as a consequence predict also those public patterns that might be impacted in reordering the private pattern. The speculation step is followed by performing reordering based on the speculated patterns, thus eliminating the processing time for reordering at the end of the window. Overall, the main goals in this thesis are as follows:

• Define a speculative decision model for reordering and implement speculative reordering strategy based on the decision model.

• Evaluate and compare speculative reordering with non-speculative-reordering.

• Document the results in a written report, present them in the department colloquium.

Requirements: The student should have good knowledge in probability theory and good programming skills (preferably C++ or python).

Refernces:

[1] Andre Soldwedel, Jutta Rothmund, Quantified Health, Technical report, YouGov Deutschland AG, 2015.

announcement as pdf