Publications
2021
1.
Biswas, Rubel; González-Castro, Víctor; Fidalgo, Eduardo; Alegre, Enrique
A new perceptual hashing method for verification and identity classification of occluded faces Artículo de revista
En: Image and Vision Computing, vol. 113, pp. 104245, 2021, (Publisher: Elsevier).
Resumen | Enlaces | BibTeX | Etiquetas: biometrics, face verification, oclussion handling, perceptual hashing
@article{biswas_new_2021,
title = {A new perceptual hashing method for verification and identity classification of occluded faces},
author = {Rubel Biswas and Víctor González-Castro and Eduardo Fidalgo and Enrique Alegre},
url = {https://www.sciencedirect.com/science/article/pii/S0262885621001505},
year = {2021},
date = {2021-01-01},
journal = {Image and Vision Computing},
volume = {113},
pages = {104245},
abstract = {Recently, research communities on Computer Vision and biometrics have shown a lot of interest in face verification and classification methods. Fighting against Child Sexual Exploitation Material (CSEM) is one of the applications that might benefit most from these advances. In CSEM, discriminative parts of the face, i.e. mostly the eyes, are often occluded to make the victim identification more difficult. Most of the current face recognition methods are not able to handle such kind of occlusions. To overcome this problem, we present One-Shot Frequency Dominant Neighborhood Structure (OSF-DNS), a new perceptual hashing method that shows advantages on two scenarios: (a) occluded face verification, i.e., matching occluded faces with their non-occluded versions, and (b) face classification, i.e., getting the identity of an occluded face by means of a classifier trained with the non-occluded faces using the perceptual hash codes as feature vectors. We have compared the face verification performance of OSF-DNS with three perceptual hashing methods and with the features obtained from five deep learning techniques, using the occluded versions of six different facial datasets. The proposed method achieves accuracies between 69.24% and 99.46% depending on the dataset, and always higher than the compared methods. For the face classification task, we compared the performance of OSF-DNS with the features obtained by four deep learning techniques. Experimental results on LFW and CFPW datasets showed that the proposed hashing method outperformed the results obtained with deep features with an accuracy up to 89.53%.},
note = {Publisher: Elsevier},
keywords = {biometrics, face verification, oclussion handling, perceptual hashing},
pubstate = {published},
tppubtype = {article}
}
Recently, research communities on Computer Vision and biometrics have shown a lot of interest in face verification and classification methods. Fighting against Child Sexual Exploitation Material (CSEM) is one of the applications that might benefit most from these advances. In CSEM, discriminative parts of the face, i.e. mostly the eyes, are often occluded to make the victim identification more difficult. Most of the current face recognition methods are not able to handle such kind of occlusions. To overcome this problem, we present One-Shot Frequency Dominant Neighborhood Structure (OSF-DNS), a new perceptual hashing method that shows advantages on two scenarios: (a) occluded face verification, i.e., matching occluded faces with their non-occluded versions, and (b) face classification, i.e., getting the identity of an occluded face by means of a classifier trained with the non-occluded faces using the perceptual hash codes as feature vectors. We have compared the face verification performance of OSF-DNS with three perceptual hashing methods and with the features obtained from five deep learning techniques, using the occluded versions of six different facial datasets. The proposed method achieves accuracies between 69.24% and 99.46% depending on the dataset, and always higher than the compared methods. For the face classification task, we compared the performance of OSF-DNS with the features obtained by four deep learning techniques. Experimental results on LFW and CFPW datasets showed that the proposed hashing method outperformed the results obtained with deep features with an accuracy up to 89.53%.
2.
Biswas, Rubel; Chaves, Deisy; Fernández-Robles, Laura; Fidalgo, Eduardo; Alegre, Enrique
A video summarization approach to speed-up the analysis of child sexual exploitation material Artículo de revista
En: XLII Jornadas de Automática, pp. 648–654, 2021, (Publisher: Universidade da Coruña, Servizo de Publicacións).
Resumen | Enlaces | BibTeX | Etiquetas: content detection, face detection, perceptual hashing, real time applications, video summarization
@article{biswas_video_2021,
title = {A video summarization approach to speed-up the analysis of child sexual exploitation material},
author = {Rubel Biswas and Deisy Chaves and Laura Fernández-Robles and Eduardo Fidalgo and Enrique Alegre},
url = {https://ruc.udc.es/dspace/handle/2183/28353},
year = {2021},
date = {2021-01-01},
journal = {XLII Jornadas de Automática},
pages = {648–654},
abstract = {This paper presents a video summarization strategy combining perceptual hashing and face detection to identify key frames from videos, specifically targeting content with faces that may relate to victims or offenders. The proposed approach is tested on adult pornography detection using the NDPI-800 dataset, achieving 84.15% accuracy and a speed of 8.05 ms/frame, making it suitable for real-time applications. This method can also create video summaries while preserving distinctive faces from the original footage.},
note = {Publisher: Universidade da Coruña, Servizo de Publicacións},
keywords = {content detection, face detection, perceptual hashing, real time applications, video summarization},
pubstate = {published},
tppubtype = {article}
}
This paper presents a video summarization strategy combining perceptual hashing and face detection to identify key frames from videos, specifically targeting content with faces that may relate to victims or offenders. The proposed approach is tested on adult pornography detection using the NDPI-800 dataset, achieving 84.15% accuracy and a speed of 8.05 ms/frame, making it suitable for real-time applications. This method can also create video summaries while preserving distinctive faces from the original footage.
2020
3.
Biswas, Rubel; González-Castro, Víctor; Fidalgo, Eduardo; Alegre, Enrique
Perceptual image hashing based on frequency dominant neighborhood structure applied to Tor domains recognition Artículo de revista
En: Neurocomputing, vol. 383, pp. 24–38, 2020, (Publisher: Elsevier).
Resumen | Enlaces | BibTeX | Etiquetas: Cybersecurity, Deep Web, perceptual hashing, TOR
@article{biswas_perceptual_2020,
title = {Perceptual image hashing based on frequency dominant neighborhood structure applied to Tor domains recognition},
author = {Rubel Biswas and Víctor González-Castro and Eduardo Fidalgo and Enrique Alegre},
url = {https://www.sciencedirect.com/science/article/pii/S0925231219316674},
year = {2020},
date = {2020-01-01},
journal = {Neurocomputing},
volume = {383},
pages = {24–38},
abstract = {This paper proposes an automatic method to recognize illicit domains on the Tor network using perceptual hashing through domain snapshots. The method introduces DUSI-2K, a dataset of Tor service domain snapshots, and F-DNS, a new hashing technique based on Dominant Neighborhood Structure (DNS) and Global Neighborhood Structure (GNS). F-DNS outperforms other state-of-the-art methods, achieving an accuracy of 98.75% in recognizing Tor domains, significantly surpassing methods like ResNet50 and Inception-ResNet-v2. Fine-tuning these models does not improve results, demonstrating the effectiveness of F-DNS for Tor domain classification.},
note = {Publisher: Elsevier},
keywords = {Cybersecurity, Deep Web, perceptual hashing, TOR},
pubstate = {published},
tppubtype = {article}
}
This paper proposes an automatic method to recognize illicit domains on the Tor network using perceptual hashing through domain snapshots. The method introduces DUSI-2K, a dataset of Tor service domain snapshots, and F-DNS, a new hashing technique based on Dominant Neighborhood Structure (DNS) and Global Neighborhood Structure (GNS). F-DNS outperforms other state-of-the-art methods, achieving an accuracy of 98.75% in recognizing Tor domains, significantly surpassing methods like ResNet50 and Inception-ResNet-v2. Fine-tuning these models does not improve results, demonstrating the effectiveness of F-DNS for Tor domain classification.
2019
4.
Biswas, Rubel; González-Castro, Víctor; Fidalgo, Eduardo; Chaves, Deisy
Boosting child abuse victim identification in Forensic Tools with hashing techniques Artículo de revista
En: V Jornadas Nacionales de Investigación en Ciberseguridad, vol. 1, pp. 344–345, 2019.
Resumen | Enlaces | BibTeX | Etiquetas: Child Sexual Abuse, CSA, Face Identification, Forensic Tools, Occluded Face Recognition, perceptual hashing
@article{biswas_boosting_2019,
title = {Boosting child abuse victim identification in Forensic Tools with hashing techniques},
author = {Rubel Biswas and Víctor González-Castro and Eduardo Fidalgo and Deisy Chaves},
url = {https://d1wqtxts1xzle7.cloudfront.net/82369096/JNIC2019_paper_11-libre.pdf?1647729671=&response-content-disposition=inline%3B+filename%3DBoosting_child_abuse_victim_identificati.pdf&Expires=1739810458&Signature=AaFk3CnTgqPwNnr4OrQ5dsZkqjASKk7CkhuGAp8EOF-QuWpTRVAHTZT4xLl1lS1NbnaDHI9p6jPc9VBhUkfmf0tLOACFOMUePTJRQbAXBLbmVv6LRjudXB1eBxP-MFeBs5vE0Ok4QaUFalGTOmM6hBXQqZC1Wb5tMxsKZPEB8eI6SxboBVqqlinMT9wyAbjykITOSmqsRSTHjOH9vHL~SZ5-PuVIbmT82w7vTdnCvovPoSwA6rmGO~IICQINI0mvYiwRYtdYVDSodTg4W7WVfst3jrcVgK76BKJHYUnH6A1HqYO38mm0MEIeXmbTSpMERXAEkc2Ja-tnkB2pyUR-Mg__&Key-Pair-Id=APKAJLOHF5GGSLRBV4ZA},
year = {2019},
date = {2019-01-01},
journal = {V Jornadas Nacionales de Investigación en Ciberseguridad},
volume = {1},
pages = {344–345},
abstract = {This work presents a new scheme for identifying occluded faces using perceptual image hashing, which differs from traditional methods by eliminating the need for prior training of a facial model. The proposed method combines frequency coefficients and statistical image information to enhance the recognition accuracy of occluded faces. The approach aims to improve face recognition in forensic applications, such as identifying victims in Child Sexual Abuse (CSA) materials. Experimental results demonstrate that the new method outperforms previous methods in occluded face identification using the LFW database.},
keywords = {Child Sexual Abuse, CSA, Face Identification, Forensic Tools, Occluded Face Recognition, perceptual hashing},
pubstate = {published},
tppubtype = {article}
}
This work presents a new scheme for identifying occluded faces using perceptual image hashing, which differs from traditional methods by eliminating the need for prior training of a facial model. The proposed method combines frequency coefficients and statistical image information to enhance the recognition accuracy of occluded faces. The approach aims to improve face recognition in forensic applications, such as identifying victims in Child Sexual Abuse (CSA) materials. Experimental results demonstrate that the new method outperforms previous methods in occluded face identification using the LFW database.
2017
5.
Biswas, Rubel; Fidalgo, Eduardo; Alegre, Enrique
Recognition of service domains on TOR dark net using perceptual hashing and image classification techniques Artículo de revista
En: 8th International Conference on Imaging for Crime Detection and Prevention (ICDP 2017), pp. 7–12, 2017, (Publisher: IET).
Resumen | Enlaces | BibTeX | Etiquetas: Darknet Detection, Image classification, perceptual hashing, TOR
@article{biswas_recognition_2017,
title = {Recognition of service domains on TOR dark net using perceptual hashing and image classification techniques},
author = {Rubel Biswas and Eduardo Fidalgo and Enrique Alegre},
url = {https://ieeexplore.ieee.org/abstract/document/8372164},
year = {2017},
date = {2017-01-01},
journal = {8th International Conference on Imaging for Crime Detection and Prevention (ICDP 2017)},
pages = {7–12},
abstract = {This paper presents a framework for identifying services on the TOR network, leveraging image content to categorize various activities such as file-sharing, ransomware, and counterfeit goods. The authors introduce the DUSI (Darknet Usage Service Images) dataset, which includes snapshots from active TOR domains across six service categories. Two pipelines were evaluated: one using Perceptual Hashing and another using Bag of Visual Words (BoVW) with SVM classifiers. The Perceptual Hashing approach achieved the highest accuracy of 99.38%, making it the recommended method for detecting TOR services based on image snapshots.},
note = {Publisher: IET},
keywords = {Darknet Detection, Image classification, perceptual hashing, TOR},
pubstate = {published},
tppubtype = {article}
}
This paper presents a framework for identifying services on the TOR network, leveraging image content to categorize various activities such as file-sharing, ransomware, and counterfeit goods. The authors introduce the DUSI (Darknet Usage Service Images) dataset, which includes snapshots from active TOR domains across six service categories. Two pipelines were evaluated: one using Perceptual Hashing and another using Bag of Visual Words (BoVW) with SVM classifiers. The Perceptual Hashing approach achieved the highest accuracy of 99.38%, making it the recommended method for detecting TOR services based on image snapshots.