Extracting evidence in forensic examinations by using deep learning methods

Today, the rapid development of technology has made it inevitable to use digital media in daily life. The fact that banks, e-government portal, e-commerce sites and social media environments make it possible to access the information very quickly has made the use of internet and digital media a constant part of daily life. This situation has led to an increase in some negative situations as well as facilitation of life. Many crimes such as fraud, data theft, the sale of illegal substances, and the spread of terrorist acts committed in digital media have become easier and have increased in number. The examination, blocking and reporting of such situations in the digital environment are the subject of digital forensics science. However, current digital forensic examination processes are unable to respond to increased crime in terms of speed. In this study, it is aimed to use the deep learning methods, which have been successful in the use of many areas, in digital evidence examinations. In this study, it is aimed to use the deep learning methods which have been successful in many use cases in judicial evidence examinations. For this purpose, a method of extracting evidence to assist in forensic examinations is proposed using deep learning object detection algorithms. The results show that the classification of images obtained from digital evidences according to their contents can be performed 93% faster using deep learning. It has also been shown that human errors will be reduced as a result of classification. It is aimed to integrate the proposed method into the forensic evidence analysis software tools by applying it to audio, text and video data as well as image data in further studies.