A Quantitative Study on Forensic Analysis of Images Produced with Deepfake Tools and Deepfake Detection

Mahmut Hilmi Baş; Ahmet Şenol

Makale detayları

A Quantitative Study on Forensic Analysis of Images Produced with Deepfake Tools and Deepfake Detection

Yazarlar:

Mahmut Hilmi Baş | Ahmet Şenol

Yayıncı:

Fırat University

Ana Başlıklar:

Adli Değerlendirme | Çıkarım ve İstatistik

Dergi:

Fırat Üniversitesi Fen Bilimleri Dergisi Vol.35 Issue 2, pp.97-118

Belge Tipi:

Makale

Yayın Tarihi:

9/2023

Belge Dili:

Türkçe

Document ID (Org No.):

2001297231

pISSN:

1308-9064

Anahtar kelimeler:

Özet:

 Todays technology pushes the limits of imagination and takes its place in a large part of our lives with fast and accessible devices. Technological growth provides great convenience to people in many areas. However, considering the speed and quantity of social media and technology reaching individuals, the impact of this technological acceleration on individuals and societies is increasing day by day. In addition to the material and moral benefits provided by social media and technology, adverse situations such as manipulated images, videos, sounds, fake news and other cyber-crimes may also be encountered. Therefore, it is important to be aware that artifacts left in the virtual world can be used by malicious individuals. This study is a methodologically quantitative study implemented in the 2022-2023 academic year. The study group of the research consists of a total of 120 participants who work in the field of computer forensics 60 participants and those who are not computer forensic experts 60 participants. The data collection tools of the research are the socio-demographic form and the True False Test developed by the researcher, which consists of 30 items to measure deepfake detection skills. According to some results of the research, the correct detection rate is lower in photographs taken through the Swapface deep editing program. It has been observed that deep editing photographs made through the Swapface program are more successful than deep editing photographs made with the Face Swapper program. It has been determined that it is not easy to detect photographs created with deep editing technology with the naked human eye and that some tools must be used.

 Todays technology pushes the limits of imagination and takes its place in a large part of our lives with fast and accessible devices. Technological growth provides great convenience to people in many areas. However, considering the speed and quantity of social media and technology reaching individuals, the impact of this technological acceleration on individuals and societies is increasing day by day. In addition to the material and moral benefits provided by social media and technology, adverse situations such as manipulated images, videos, sounds, fake news and other cyber-crimes may also be encountered. Therefore, it is important to be aware that artifacts left in the virtual world can be used by malicious individuals. This study is a methodologically quantitative study implemented in the 2022-2023 academic year. The study group of the research consists of a total of 120 participants who work in the field of computer forensics 60 participants and those who are not computer forensic experts 60 participants. The data collection tools of the research are the socio-demographic form and the True False Test developed by the researcher, which consists of 30 items to measure deepfake detection skills. According to some results of the research, the correct detection rate is lower in photographs taken through the Swapface deep editing program. It has been observed that deep editing photographs made through the Swapface program are more successful than deep editing photographs made with the Face Swapper program. It has been determined that it is not easy to detect photographs created with deep editing technology with the naked human eye and that some tools must be used.

 Todays technology pushes the limits of imagination and takes its place in a large part of our lives with fast and accessible devices. Technological growth provides great convenience to people in many areas. However, considering the speed and quantity of social media and technology reaching individuals, the impact of this technological acceleration on individuals and societies is increasing day by day. In addition to the material and moral benefits provided by social media and technology, adverse situations such as manipulated images, videos, sounds, fake news and other cyber-crimes may also be encountered. Therefore, it is important to be aware that artifacts left in the virtual world can be used by malicious individuals. This study is a methodologically quantitative study implemented in the 2022-2023 academic year. The study group of the research consists of a total of 120 participants who work in the field of computer forensics 60 participants and those who are not computer forensic experts 60 participants. The data collection tools of the research are the socio-demographic form and the True False Test developed by the researcher, which consists of 30 items to measure deepfake detection skills. According to some results of the research, the correct detection rate is lower in photographs taken through the Swapface deep editing program. It has been observed that deep editing photographs made through the Swapface program are more successful than deep editing photographs made with the Face Swapper program. It has been determined that it is not easy to detect photographs created with deep editing technology with the naked human eye and that some tools must be used.

 Günümüz teknolojisi hayal gücünün sınırlarını zorlayarak hızlı ve ulaşılabilir cihazlarla yaşantımızın büyük bir bölümünde yerini almaktadır. Teknolojik büyüme birçok alanda insanlara büyük kolaylıklar sağlamaktadır. Ancak, sosyal medyanın ve teknolojinin bireylere ulaşma hızı ve niceliği göz önüne alındığında, bu teknolojik ivmenin bireyler ve toplumlar üzerindeki etkisi her geçen gün artmaktadır. Sosyal medya ve teknolojinin sağladığı maddi ve manevi faydaların yanı sıra, manipüle edilmiş resimler, videolar, sesler, sahte haberler ve diğer siber suçlar gibi aksi durumlarla da karşılaşılabilmektedir. Bu nedenle, sanal dünyada bırakılan kalıntıların kötü niyetli kişiler tarafından kullanılabileceği konusunda bilinçli olmak önemlidir. Bu çalışma, 2022-2023 eğitim-öğretim yılında uygulanmış, metodolojik açıdan nicel bir çalışmadır. Araştırmanın çalışma grubu, adli bilişim alanında çalışan 60 katılımcı ve adli bilişimci olmayan 60 katılımcı toplam 120 katılımcıdan oluşmaktadır. Araştırmanın veri toplama araçları, sosyo-demografik form ve araştırmacı tarafından geliştirilen ve derin kurgu deepfake tespit becerisini ölçmek için 30 maddeden oluşan Doğru Yanlış Testidir. Araştırmanın bazı sonuçlarına göre, Swapface derin kurgu yapma programı vasıtasıyla yapılan fotoğraflarda doğru tespit oranı daha düşüktür. Swapface programı vasıtasıyla yapılan derin kurgu fotoğraflarının, Face Swapper programıyla yapılan derin kurgu fotoğraflarına göre daha başarılı olduğu görülmüştür. Derin kurgu teknolojisiyle oluşturulan fotoğrafların tespit edilmesinde çıplak insan gözüyle tespitin kolay olmadığı, birtakım araçların kullanılması gerektiği belirlenmiştir.

Makale detayları

A Quantitative Study on Forensic Analysis of Images Produced with Deepfake Tools and Deepfake Detection

Yazarlar:

Mahmut Hilmi Baş | Ahmet Şenol

Yayıncı:

Fırat University

Ana Başlıklar:

Adli Değerlendirme | Çıkarım ve İstatistik

Dergi:

Fırat Üniversitesi Fen Bilimleri Dergisi Vol.35 Issue 2, pp.97-118

Belge Tipi:

Makale

Yayın Tarihi:

9/2023

Belge Dili:

Türkçe

Document ID (Org No.):

2001297231

pISSN:

1308-9064

Forensics

deepfake

machine learning

manipulation

cyber security

artificial intelligenceAdli bilişim

deepfake

makine öğrenmesi

manipülasyon

siber güvenlik

yapay zekâ.

Sayı

Makale detayları

Yazarlar: Mahmut Hilmi Baş | Ahmet Şenol

Yayıncı: Fırat University

Ana Başlıklar: Adli Değerlendirme | Çıkarım ve İstatistik

Dergi: Fırat Üniversitesi Fen Bilimleri Dergisi Vol.35 Issue 2, pp.97-118

Belge Tipi: Makale

Yayın Tarihi: 9/2023

Belge Dili: Türkçe

Document ID (Org No.): 2001297231

pISSN: 1308-9064

Sayı

Yazarlar:

Mahmut Hilmi Baş | Ahmet Şenol

Yayıncı:

Fırat University

Ana Başlıklar:

Adli Değerlendirme | Çıkarım ve İstatistik

Dergi:

Fırat Üniversitesi Fen Bilimleri Dergisi Vol.35 Issue 2, pp.97-118

Belge Tipi:

Makale

Yayın Tarihi:

9/2023

Belge Dili:

Türkçe

Document ID (Org No.):

2001297231

pISSN:

1308-9064