Search

The development of information technologies, including artificial intelligence, is leading to their widespread use in many areas of our lives. With the rapid development of artificial intelligence, the number of cases of its application to generate digital images, audio, video, etc. is growing. Counterfeiting digital audio files is dangerous from the point of view of being used by criminals to exert information and psychological influence and manipulate society and individuals. There are modern methods for detecting audio fakes made by artificial intelligence. They are highly accurate, but not without drawbacks. The main drawback is the complex architecture and high resource intensity. The aim of this paper is to develop a neural network that would allow detecting audio fakes with satisfactory accuracy and whose training would not require significant computing resources. A method was chosen for modification, namely, the method of detecting audio fakes using a convolutional neural network. The method for detecting audio fakes was modified by building a model with a new architecture with fewer layers, which significantly reduced the time and computational resources required compared to analogues. The experimental results obtained using the modified method show satisfactory efficiency and accuracy of the system. Errors of the first kind amounted to 24%, errors of the second kind - 9%. The developed modification is implemented in a software application with a convenient and simple interface. [ABSTRACT FROM AUTHOR]

Methods of detecting violations of the integrity of digital information (digital photo, video, and audio files) play an important role for information and cyber security. Among the violations of the integrity of digital images, post-processing after possible falsification through cloning or photomontage can be singled out as a separate category. Post-processing can be done with various tools: blurring, sharpening, multiplicative noise, etc. The methods of detecting multiplicative noise are poorly illuminated in open sources. There is a method based on the analysis of the DCP coefficients of the digital image matrix, which has proven itself well when saving processed files without loss. However, it has not been tested when saved in a lossy format. The purpose of the work is to increase the effectiveness of multiplicative noise detection as a digital image processing. The computational experiment was carried out using a database of 100 images, each of which was subjected to multiplicative noise processing with a variance of 0.005 and saved in jpeg as a quality from 0 to 100 with a step of 5. The difference between the maximum and minimum value of high frequencies (amplitude) was found in two different ways and two indicators are defined. It was established that the accuracy of multiplicative noise detection depends on the degree of lossy compression. The accuracy of noise detection is greatly reduced when the compression ratio is set below 70. The P1 indicator is suitable for detecting multiplicative noise in images with quality from 98 to 100 and from 70 to 89, P2 - for images with quality from 89 to 97. The conducted research allowed to significantly improve the effectiveness of detecting multiplicative noise as a violation of the integrity of a digital image. [ABSTRACT FROM AUTHOR]

The information technologies are spreading in the modern world and are one or more interconnected software products for a particular computer that helps the user achieve this aim. Every day we receive a wealth of digital information, such as emails, articles on the Internet, messages in various communication systems, and more. Therefore, the issue of qualitative and rapid verification of its authenticity, namely, elements such as digital images, is very timeliness. Digital images are often not original. They have different types of integrity violations, such as cloning or collage. After applying these operations, the different filters for blurring, brightness and contrast variation, sharpening, etc. often are used to hide the digital image processing .These operations are performed by graphic editors such as GIMP. In this graphic editor, which is a free analogue to the Adobe Photoshop graphic editor, sharpening is implemented as an Unsharp Mask Filter. A review of open source literature has shown that little attention is paid to detecting artificial sharpening. From open sources it is known about the method for detection of artificial sharpening based on the analysis of close color pairs of the image matrix. Its main disadvantage is the high number of the type I errors and the type II errors. This fact leads to the development of new methods and algorithms for detection this type of image processing. The object of this work is to detect artificial sharpening of a digital image as an integrity violation. A study is conducted in which the formal parameters of the digital image matrix were revealed, indicating the presence of artificial sharpening. A quantitative estimation of the qualitative differences between processed and raw images is obtained and used as a threshold for detecting artificially sharpened images. An algorithm for the detection of the artificial sharpening of a digital image is developed and its efficiency is estimated, According to this algorithm, the number of the type I errors is 3,8%, and the number of the type II errors is 9,8%. The development potential of this work is to improve the detection of different ways of artificial sharpening the digital image. [ABSTRACT FROM AUTHOR]

The use of the digital images in modern society is very common, making it difficult to name an industry where digital images are not used. However, in certain situations, the digital images may be direct or indirect digital evidence, such as in court cases. Tampering or violating the integrity of the digital images is one of the important issues in the information protection. The issue of the digital signal authentication by cybersecurity experts is a pressing issue in the present days. Therefore, the development of methods and algorithms for detecting the integrity of digital signals in general, and digital images in particular, is a very crucial task. Among all the possible variations in the integrity violation of digital images can be distinguished a separate group, in which the digital images are processed by various filters editors, such as blurring, sharpening, etc. The blurring is often applied to images not only when they are tampered, but also as a steganographic attack. Blur detection indicates the inability to use a digital image as the evidence of anything. One of the most effective blur detection methods known from the open sources is a method based on the analysis of singular values of blocks of a digital image matrix. The aim of this paper is to increase the efficiency of the blur detection in a digital image by modifying an algorithm based on the analysis of singular values. Computational experiments are performed to determine the number of parameters to be verified. The results obtained allow modifying the algorithm of the method of the blur detection in a digital image. The modification makes it possible to detect tampered access to a digital image. The modified algorithm is no less effective than the original algorithm, and allows analyzing fewer parameters in a shorter time. The computational complexity of the algorithm is determined by the second order polynomial that is acceptable in terms of the practical application. [ABSTRACT FROM AUTHOR]

Due to the growth of computer crime, an important and urgent problem is the development of methods for detecting the integrity violations of a digital image. In particular, the detection of different types of the image post-processing after their possible forgery is a grey area of information security, and the detection of processing by the Sharpen Filter is an issue practically not covered in open scientific literature. The artificial sharpening of a digital image is the reverse procedure to the blurring. Previously, the authors of the paper developed the method for detecting blur of a digital image based on an analysis of the growth rate of singular values. When blurring, the growth rate of singular values is an order of magnitude smaller than before the blurring, which makes it possible to reveal a threshold value for detecting blurred images. The singular values increase accordingly to the artificially sharpening. But it is not possible to reveal the threshold value in this case. The aim of this paper is to identify the parameters of the digital image matrix, which indicate an artificial sharpening, and to develop a method for detecting sharpness as a digital image post-processing. If the contour is considered as a change in the brightness function of a digital image, then the change in the brightness function will be faster when the contours are sharpened. If two adjacent pixels had the close sharpness values before sharpening, then applying this filter will increase the difference between them. In this paper, the investigation of the close color pairs of an image is carried out when the Sharpness Filter was applied to it. It is revealed that the number of close color pairs during processing decreases. The threshold value which allows separating the processed image from the raw one is empirically established. The method of detecting sharpness as a digital image postprocessing is developed based on the results obtained. [ABSTRACT FROM AUTHOR]

In the work the modification of a method of detection and localization of regions of the integrity violation in the digital image obtained as a result of cloning is developed. Aim: The aim of the work is to increase the accuracy of detection of integrity violation regions of the digital image by development of modification of a method of detection and localization of cloning regions based on the block oriented approach due to simultaneous use in case of expertize of the image with square, triangular, round blocks and of irregular shape blocks. Materials and Methods: Further development of the block oriented approach to the solution of problem of cloning results detection in digital image is provided. Modification of the method of detection and localization of the cloning regions developed by authors earlier, based on the accounting of closeness of the matrix blocks which is quantitatively evaluated by means of different variants of the metrics assignment is offered. Results: The technique of simultaneous use of square, triangular, round blocks and blocks of irregular shape in case of the expertize of the image based on the block oriented approach is developed. Providing the computational experiment the accuracy increase of determination of integrity violation regions of the digital image when using of the offered modification of the method of detection and localization of fields of cloning in case of simultaneous use in case of expertize of different partitioning types blocks is set: the relative value of the area of the revealed region from real region of a clone increases by 1.6 times in comparison with results of operation of algorithmic implementation of the method taken as a basis. [ABSTRACT FROM AUTHOR]

The widespread using of fake pictures in information wars, lawsuits, politics, etc., needs to develop new detection methods of photomontage. One of them is a method of photomontage detection based on the analysis of singular values of matrix blocks of digital images. Aim: The aim of this research is to check the application area of photomontage detection method based on analysis of singular values of matrix blocks of digital images in the absence of restrictions on the degree of their compression. Materials and Methods: The basis of this method is the detection of the differences between the images when they have different degrees of compression: values of the zero singular values matrix of blocks corresponding to foreign part of image. We distinguish this area on the background of the main image, that allows to detect falsification. The mentioned method was investigated under certain restrictions imposed on the used test photos. An advanced experiment with images was held, for which there are no restrictions on the degree of compression. Results: The falsification was detected in 44.4% of cases in experimental study while saving the digital image after photomontage in lossless format, and in 53.3% of cases when saving the losses. Montage was detected most often when the difference between the quality factor Q of the main image and the replacing area was more than two. [ABSTRACT FROM AUTHOR]

Developing methods aiming at identification of digital images integrity violation is an important problem of complex system of information security. Cloning method, realized in the most of graphic editors, is often used to compromise the integrity of the image related to unauthorized changes. Therefore, the task of cloning identification is a relevant one. In addition to the cloning identification it is also important to clarify the contours of the cloned area. In this paper we propose a polynomial comprehensive method to identify areas of cloning as a falsification of the digital image. It allows to locate the cloned areas with a higher precision compared to modern analogues. Complex method is based on two modern clone detection techniques, using a common approach to the analysis of the state and technology of the information system functioning. It includes two stages: "fast rough" cloning area identification and a "slow" identification of the contour of the cloning area using a limited image subarea containing a clone. The method proposed can be adapted for the digital video. [ABSTRACT FROM AUTHOR]