東華大學圖書館 |

A Machine Learning Based Digital Forensics Application to Detect Tampered Multimedia Files.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	A Machine Learning Based Digital Forensics Application to Detect Tampered Multimedia Files./
作者:	Ferreira, Sara Cardoso.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2021,
面頁冊數:	104 p.
附註:	Source: Dissertations Abstracts International, Volume: 84-01, Section: B.
Contained By:	Dissertations Abstracts International84-01B.
標題:	User interface. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=29150422
ISBN:	9798835567041

A Machine Learning Based Digital Forensics Application to Detect Tampered Multimedia Files.
Ferreira, Sara Cardoso.

A Machine Learning Based Digital Forensics Application to Detect Tampered Multimedia Files. - Ann Arbor : ProQuest Dissertations & Theses, 2021 - 104 p.

Source: Dissertations Abstracts International, Volume: 84-01, Section: B.

Thesis (M.E.)--Universidade do Porto (Portugal), 2021.

This item must not be sold to any third party vendors.

Computer Forensics, commonly known as Digital Forensics, embodies techniques, tools, and procedures for the collection, preservation and analysis of digital evidence in electronic equipment, including disks, smartphones and other devices with storage capacity. The artifacts collected and with interest for the criminal investigation are diverse as depend on the crime being investigated. Besides the common artifacts, there are several types of digital content that may have potential interest to the investigator, such as files, e-mail addresses, phone contacts or other information that may indicate the existence of criminal activity.Multimedia content (photos and videos) have a high interest to digital forensics, mainly due to its association with crimes that have a high impact on public opinion. Fake news, misinformation, sexual abuse, and distribution of pornographic content, especially involving under aged, are some examples where seized videos and photos are relevant to the investigation and must necessarily be analyzed by criminal investigation teams. This analysis, usually employs manual and time-consuming techniques.There are several tools and techniques used to process and analyze photos and videos, with numerous applications in facial recognition and emotion detection. Although advances can also be identified in the detection of manipulated multimedia content, this has not been translated into improvements for forensic analysis and for the investigation of cybercrimes. Autopsy is a fast, easy-to-use, and open source digital forensic platform, which is capable of analyzing a wide range of electronic devices. Besides the built-in modules, it also benefits from the developments made by the community, mainly through the development of custom modules.This dissertation describes the development of a Support Vector Machines (SVM)-based standalone application and two modules for Autopsy, to automate photos and videos files processing, and to compute their probability of have been digitally manipulated. A dataset comprising both real and manipulated files containing objects and faces, was also created to train and test the model. Using a ten-fold cross-validation it was achieved an F1-score above 99.5% for tampered photos detection, 79.8% for videos, and 89.2% for a mixture of manipulated photos and videos. The method used was compared with a Convolutional Neural Network (CNN)-based method, to benchmark both models. Although investigators' manual analysis is still needed, the developed modules indicate the examples that should deserve the investigator's attention, that is those that have a higher probability of have been manipulated.

ISBN: 9798835567041Subjects--Topical Terms:

3681528
User interface.

A Machine Learning Based Digital Forensics Application to Detect Tampered Multimedia Files.
LDR:06520nmm a2200349 4500 001 2352228
005 20221118093849.5
008 241004s2021 ||||||||||||||||| ||eng d
020 $a 9798835567041
035 $a (MiAaPQ)AAI29150422
035 $a (MiAaPQ)Portugal10216135823
035 $a AAI29150422
040 $a MiAaPQ $c MiAaPQ
100 1 $a Ferreira, Sara Cardoso. $3 3691849
245 1 0 $a A Machine Learning Based Digital Forensics Application to Detect Tampered Multimedia Files.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2021
300 $a 104 p.
500 $a Source: Dissertations Abstracts International, Volume: 84-01, Section: B.
500 $a Advisor: Correia, Manuel Eduardo Carvalho Duarte;Antunes, Mario Joao Goncalves.
502 $a Thesis (M.E.)--Universidade do Porto (Portugal), 2021.
506 $a This item must not be sold to any third party vendors.
520 $a Computer Forensics, commonly known as Digital Forensics, embodies techniques, tools, and procedures for the collection, preservation and analysis of digital evidence in electronic equipment, including disks, smartphones and other devices with storage capacity. The artifacts collected and with interest for the criminal investigation are diverse as depend on the crime being investigated. Besides the common artifacts, there are several types of digital content that may have potential interest to the investigator, such as files, e-mail addresses, phone contacts or other information that may indicate the existence of criminal activity.Multimedia content (photos and videos) have a high interest to digital forensics, mainly due to its association with crimes that have a high impact on public opinion. Fake news, misinformation, sexual abuse, and distribution of pornographic content, especially involving under aged, are some examples where seized videos and photos are relevant to the investigation and must necessarily be analyzed by criminal investigation teams. This analysis, usually employs manual and time-consuming techniques.There are several tools and techniques used to process and analyze photos and videos, with numerous applications in facial recognition and emotion detection. Although advances can also be identified in the detection of manipulated multimedia content, this has not been translated into improvements for forensic analysis and for the investigation of cybercrimes. Autopsy is a fast, easy-to-use, and open source digital forensic platform, which is capable of analyzing a wide range of electronic devices. Besides the built-in modules, it also benefits from the developments made by the community, mainly through the development of custom modules.This dissertation describes the development of a Support Vector Machines (SVM)-based standalone application and two modules for Autopsy, to automate photos and videos files processing, and to compute their probability of have been digitally manipulated. A dataset comprising both real and manipulated files containing objects and faces, was also created to train and test the model. Using a ten-fold cross-validation it was achieved an F1-score above 99.5% for tampered photos detection, 79.8% for videos, and 89.2% for a mixture of manipulated photos and videos. The method used was compared with a Convolutional Neural Network (CNN)-based method, to benchmark both models. Although investigators' manual analysis is still needed, the developed modules indicate the examples that should deserve the investigator's attention, that is those that have a higher probability of have been manipulated.
520 $a A Informatica Forense, vulgarmente designada por Analise Digital Forense, emprega tecnicas, aplicacoes e procedimentos para a recolha, preservacao e analise de evidencias digitais em equipamentos electronicos, nomeadamente discos, smartphones e outros dispositivos com capacidade de armazenamento. Os artefactos recolhidos e de interesse para a investigacao criminal sao diversos, dependendo do crime em causa. Para alem dos artefactos comuns, ha outros conteudos digitais que podem ter potencial interesse para o investigador, como ficheiros, enderecos de e-mail, contactos e outra informacao que pode indicar a existencia de aividade criminal.Os conteudos multimedia (fotos e videos) tem um interesse acrescido pela sua associacao a crimes com um elevado impacto na opiniao publica. As noticias falsas, o abuso sexual e a distribuicao de conteudo pornografico, especialmente envolvendo menores, sao alguns exemplos onde as fotos e videos apreendidos sao relevantes para a investigacao e tem de ser analisados pelas equipas de investigacao criminal. A analise deste tipo de conteudos recorre normalmente a tecnicas manuais e pouco sistematizadas.Existem varias aplicacoes e tecnicas para processar e analisar fotos e videos, com inumeras aplicacoes no reconhecimento facial e na detecao de emocoes. Embora se possam identificar avancos tambem na detecao de conteudos multimedia manipulados, tal nao se tem traduzido em melhorias para a informatica forense e para a investigacao de crimes informaticos. Autopsy e uma plataforma de analise digital forense open-source, rapida, facil de usar e capaz de analisar uma vasta gama de dispositivos. Alem dos modulos nativos, o Autopsy beneficia tambem do desenvolvimento feito pela comunidade, nomeadamente pela adicao de modulos personalizados.Esta dissertacao descreve uma aplicacao baseada em Support Vector Machines (SVM) e de dois modulos para o Autopsy, para automatizar o processamento de fotos e videos, e calcular a probabilidade de terem sido manipulados digitalmente. Foi tambem criado um dataset contendo fotos e videos reais e manipulados, com objectos e faces, para treinar e testar o modelo. Os testes realizados com ten-fold cross validation, revelaram um F1 acima de 99, 5% na detecao de fotos manipuladas, 79, 8% na detecao de videos, e 89, 2% na detecao de uma mistura de fotos e videos. O metodo utilizado foi comparado com um metodo baseado em Convolutional Neural Network (CNN), para avaliar ambos os modelos. Embora a analise manual ainda seja necessaria, os modulos desenvolvidos indicam os ficheiros que devem merecer maior atencao pelo investigador, ou seja, os que tem maior probabilidade de terem sido manipulados.
590 $a School code: 5896.
650 4 $a User interface. $3 3681528
650 4 $a Internet crime. $3 3541387
650 4 $a Support vector machines. $3 2058743
650 4 $a Artificial intelligence. $3 516317
650 4 $a Computer science. $3 523869
650 4 $a Criminology. $3 533274
650 4 $a Web studies. $3 2122754
690 $a 0800
690 $a 0984
690 $a 0627
690 $a 0646
710 2 $a Universidade do Porto (Portugal). $3 3348738
773 0 $t Dissertations Abstracts International $g 84-01B.
790 $a 5896
791 $a M.E.
792 $a 2021
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=29150422