PerSec – A Privacy Preserving and Trustworthy Middleware for Mobility and Energy Applications

We are building ‘PerSec’ – a middleware system providing homomorphic encryption and state of the art anomaly detection services. The middleware also provides the ability to fine tune these services at runtime while managing the application lifecyles. We focus on mobility and energy applications as these are two critical infrastructure of our communities.

Context

The rapid evolution of data-driven analytics, Internet of things (IoT) and cyber-physical systems (CPS) are fueling a growing set of Smart and Connected Communities (SCC) applications, including for smart transportation and smart energy. However, the deployment of such technological solutions without proper security mechanisms makes them susceptible to data integrity and privacy attacks, as observed in a large number of recent incidents. The goal of this project is to develop a framework to ensure data privacy, data integrity, and trustworthiness in smart and connected communities. The innovativeness of the project lies in the collaborative effort between team of researchers from US and Japan together. As part of the project the research team is developing privacy-preserving algorithms and models for anomaly detection, trust and reputation scoring used by application providers for data integrity and information assurance. Towards that goal, we are also studying trade-offs between security, privacy, trust levels, resources, and performance using two exemplar applications in smart mobility and smart energy exchange in communities.

Research

We conduct research across five thrusts.

  • Encryption
  • Anomaly Detection
  • Novel Applications
  • Runtime QoS Configuration Middleware
  • System Integration

Significant Results

  • We proposed a novel framework for privacy-preserving anomaly-based data falsification attack detection over fully homomorphic encryption (FHE) data for smart metering infrastructure. Unlike existing privacy-preserving anomaly detectors, this framework detects not only energy theft but also more advanced data integrity attacks.
  • We optimized the anomaly detection algorithm for computational efficiency, thus making it practical for resource-constrained devices such as smart meters, achieving a 40x speed-up over the well known CKKS FHE method.
  • We validated our proposed framework with real datasets from smart metering infrastructures, and demonstrated that the data integrity attacks can be detected with high sensitivity yet without sacrificing user privacy.
  • With a real smart meter dataset (one hour resolution) of 200 houses from Texas and the Irish dataset we were able to demonstrate that the detection sensitivity of the plain-text anomaly identification algorithm is not degraded due to the use of homomorphic encryption.
  • We developed a decentralized city level middleware for mobility applications using roadside units (RSUs) as edge processing devices and developed a Decentralized Route Planning with integrated anomaly detection. For this purpose, we divided the city road network into grids, each assigned an RSU where the traffic data are kept locally, thus increasing security and resiliency such that the system can perform even if some RSUs fail. Route generation is done in two steps. First, an optimal grid sequence is generated, prioritizing shortest path calculation accuracy but not RSU load. Second, we assign route planning tasks to the grids in the sequence. Keeping in mind the RSU load and constraints, tasks can be allocated and executed in any non-optimal grid but with lower accuracy. We evaluate this system using Metropolitan Nashville road traffic data. We divided the area into 500 grids, configuring load and neighborhood sizes to meet delay constraints while maximizing model accuracy. Results show a 30% decrease in processing time with a decrease in model accuracy of 99% to 92.3%, by simply increasing the search area to the optimal grid’s immediate neighborhood.
  • We extended the middleware to enable federated learning to collaboratively learn shared prediction models online. The middleware is currently being used as the baseline to integrate the homomorphic encryption modules. We are using it to study trade-offs under heterogeneous networking and computational resource constraints.

Investigators

Acknowledgement

This material is based upon work supported by the National Science Foundation under Grant No. 1818901 and 1818942 and Japan Science and Technology.

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.”