Publication: Identificación de la Fuente de Adquisición de Imágenes/Vídeos en Escenarios Abiertos usando Técnicas de Inteligencia Artifical
Loading...
Official URL
Full text at PDC
Publication Date
2018
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
El creciente uso de nuevas tecnologías en las últimas décadas ha llevado a que los teléfonos móviles sean los dispositivos más utilizados a diario. Como consecuencia, todos los días se toman millones de imágenes y vídeos, generando una gran cantidad de información que puede ser utilizada como evidencia en tribunales. Por tanto, es completamente necesario introducir mecanismos para garantizar la identificación de la fuente de lo que puede ser considerado como prueba en procesos judiciales. En este trabajo se propone un algoritmo para la identificación de la fuente de vídeos, basado en imperfecciones inherentes que los sensores presentan debido al proceso de fabricación, lo que permite distinguir entre dos móviles del mismo modelo y marca. Estas imperfecciones son extraídas de los fotogramas relevantes de los vídeos a través la transformada de ondícula de Daubechies y mediante agrupamiento los vídeos son asociados en distintas clases basados en la correlación. Para determinar el número óptimo de grupos se utiliza el método del codo.
The increasing use of new technologies over the last few decades has come with mobile phones being the most employed device on a daily basis. As a consequence, millions of images and videos are generated every day, originating a huge amount of information that could possibly be used as evidence in court. Therefore, it is completely necessary to introduce procedures that guarantee the identification of the origin from what can be considered proof in forensic matters. In this work an algorithm to identify the source of a video is proposed, based on inherent imperfections that each sensor present due to the manufacturing process, which allows to distinguish between two different mobiles from the same model and brand. These imperfections are extracted from the relevant frames from the videos with Daubechies Wavelet Transform and by means of clustering are grouped into different classes based on their correlation. The elbow method is used to identify the optimal number of clusters.
The increasing use of new technologies over the last few decades has come with mobile phones being the most employed device on a daily basis. As a consequence, millions of images and videos are generated every day, originating a huge amount of information that could possibly be used as evidence in court. Therefore, it is completely necessary to introduce procedures that guarantee the identification of the origin from what can be considered proof in forensic matters. In this work an algorithm to identify the source of a video is proposed, based on inherent imperfections that each sensor present due to the manufacturing process, which allows to distinguish between two different mobiles from the same model and brand. These imperfections are extracted from the relevant frames from the videos with Daubechies Wavelet Transform and by means of clustering are grouped into different classes based on their correlation. The elbow method is used to identify the optimal number of clusters.
Description
Trabajo de Fin de Máster, Universidad Complutense, Facultad de Informática, Departamento de Ingeniería del Software e Inteligencia Artificial, Curso 2017/2018