Thiago Moreira da Costa - An ontology-based privacy preservation approach for the Internet of Things

Organized by: 
Thiago Moreira da Costa
Thiago Moreira da Costa
Detailed information: 


Venue :
Room  206 in the  IMAG building (Université Grenoble Alpes, 700, avenue centrale, 38401 Saint Martin d'Hères).

Thesis Defense Committee Members :

Thesis supervisor :
Hervé MARTIN, Prof. at the Université Grenoble Alpes
Thesis co-supervisor :
Nazim AGOULMINE, Prof. at the Université d'Évry Val-d'Ossonnes - Paris

Referees :
Maryline LAURENT, Prof.  at the Institut Mines-Télécom, TELECOM Sud Paris
Karine BENNIS ZEITOUNI - Prof.  at the Université de Versailles St. Quentin - Versailles

Inspectors :
Reinaldo BEZERRA BRAGA - Prof. at the Federal Institut of Ceará - Aracati - Brazil
Didier DONZES - Prof. at the Université Grenoble Alpes



The spread of pervasive computing through the Internet of Things (IoT) represents a challenge for privacy preservation. Privacy threats are directly related to the capacity of the IoT sensing to track individuals in almost every situation of their lives. Allied to that, data mining techniques have evolved and has been used to extract a myriad of personal information from sensor data stream. This trust model relies on the trustworthiness of the data consumer who should infer only intended information. However, this model exposes personal information to privacy adversary. In order to provide a privacy preservation for the IoT, we propose a privacy-aware virtual sensor model that enforces privacy policy in the IoT sensing. This mechanism intermediates physical sensors and data consumers. As a consequence, we are able to optimize the use of privacy preserving techniques by applying them selectively according to virtual sensor inference intentions, while preventing malicious virtual sensors to execute or get direct access to raw sensor data. In addition, we propose an ontology to classify personal information based on the Behavior Computing, facilitating privacy policy definition and information classification based on the behavioral contexts.