Inproceedings
Assessing the Influencing Factors on the Accuracy of Underage Facial Age Estimation
Contribution Summary
This research assesses the factors influencing the accuracy of automated underage facial age estimation using two cloud services, Microsoft Azure's Face API and Amazon Web Service's Rekognition service. The study evaluates the performance of these services on the VisAGe dataset, a comprehensive collection of underage facial images. The analysis reveals correlations between facial attributes, such as facial expressions, noise, makeup, and image quality, and age estimation errors. The results identify the most significant factors affecting the accuracy of underage facial age estimation, providing insights for future age estimation modeling and development of more accurate automated age estimation systems.
Keywords: Automated facial age estimation; Digital forensics; Cloud services; Facial attributes; Age estimation errors; VisAGe dataset; Microsoft Azure Face API; Amazon Web Service Rekognition service
Abstract
Swift response to the detection of endangered minors is an ongoing concern for law enforcement with the rapid growth of disk capacities and data being stored in the cloud. Automated tools are needed to aid in CSEM investigation - both to expedite the evidence discovery process, while lessening the investigator's exposure to traumatic material. In these investigations, age estimation techniques show great promise in helping decrease the overflowing backlog of evidence obtained from the vast array of devices and online services. A lack of sufficient training data combined with natural human variance has been hindering accurate automated age estimation, especially for underage subjects. A comprehensive evaluation of the performance achieved on over 21,800 underage subjects with two cloud age estimation services is presented, namely Amazon Web Service's Rekognition service and Microsoft Azure's Face API. The objective of this work is to evaluate the influence that certain human biometric factors, facial expressions, and image quality, i.e., blur, noise, exposure and resolution, have on the outcome of automated age estimation services. The thorough evaluation of the correlation and effects of such factors aids the understanding of the performance and allows us to identify the most influencing factors to be overcome in future age estimation modelling.
BibTeX
@inproceedings{anda2020InfluencingFactorsAgeEstimation,
author={Anda, Felix and Becker, Brett and Lillis, David and Le-Khac, Nhien-An and Scanlon, Mark},
title="{Assessing the Influencing Factors on the Accuracy of Underage Facial Age Estimation}",
booktitle="{The 6th IEEE International Conference on Cyber Security and Protection of Digital Services (Cyber Security)}",
year=2020,
month=06,
location={Virtual Event},
publisher={IEEE},
abstract="Swift response to the detection of endangered minors is an ongoing concern for law enforcement with the rapid growth of disk capacities and data being stored in the cloud. Automated tools are needed to aid in CSEM investigation -- both to expedite the evidence discovery process, while lessening the investigator's exposure to traumatic material. In these investigations, age estimation techniques show great promise in helping decrease the overflowing backlog of evidence obtained from the vast array of devices and online services. A lack of sufficient training data combined with natural human variance has been hindering accurate automated age estimation, especially for underage subjects. A comprehensive evaluation of the performance achieved on over 21,800 underage subjects with two cloud age estimation services is presented, namely Amazon Web Service's Rekognition service and Microsoft Azure's Face API. The objective of this work is to evaluate the influence that certain human biometric factors, facial expressions, and image quality, i.e., blur, noise, exposure and resolution, have on the outcome of automated age estimation services. The thorough evaluation of the correlation and effects of such factors aids the understanding of the performance and allows us to identify the most influencing factors to be overcome in future age estimation modelling.",
doi={10.1109/CyberSecurity49315.2020.9138851},
}