CC BY
0DETECTION OF FAKE INFORMATION
Shoakhmedova Nozima Khairullaevna
Associate Professor, Department of Digital Economy, TDIU (Uzbekistan). https://doi.org/10.5281/zenodo.11210752
Abstract. This article is devoted to determining the reliability of information and recognizing fake content. Information about the use of artificial intelligence methods and technologies is covered. Explains methods for distinguishing false information from real information in information systems. Reveals processes using artificial intelligence to determine whether information in images or text has changed.
Key words: artificial intelligence, neural netwоrks, security system, health check, оptical character recоgnitiоn (OCR), machine readable zоne ^de check.
INTRODUCTION
Artificial intelligence (AI) is a specific branch of computer science that deals with the creation of computer systems with capabilities typically associated with the human mind: language understanding, learning, discussion, problem solving, translation, and similar capabilities.
Currently, AI consists of algorithms and software packages designed to perform various actions and can cope with several tasks that the human mind can do.
One striking example is the IBM Watson supercomputer, which, based on its database, answers questions in a specific language. Also, one of SI's achievements includes programs such as Siri, a mobile assistant that has become a constant companion for many people, and Prisma, a photo processor. By now, artificial intelligence has become widely popular and covers almost every aspect of our daily lives. For example, residents of the Chinese city of Yingchuan do not need bank cards. All processes related to calculations are carried out by artificial intelligence by identifying a person's face[1].
Artificial intelligence is a set of hardware and software tools that allow a non-programming user to create and solve their own problems.
The problem of creating artificial intelligence aroused people's interest long before the appearance of the first computers. But today the human brain is more efficient than any supercomputer. All modern computers are built in such a way that the computer and storage space are separated from each other. In the brain, computing and data storage are done in one place, allowing speed, energy consumption, and information processing to begin.
Until now, no system has been able to recognize visual and auditory images like a living brain. Recognition is a process associated with searching for information in a database, comparing the image seen with already familiar ones, identifying common and unique features, etc[2].
Nowadays, AI is helping us a lot in identifying fake documents, which can be used to prevent falsification of information.
The most popular use cases for falsifying information are:
- Finance (banks, payment systems and credit companies);
- Government bodies;
- Human Resources Department;
- Travel (verification of identity and health documents, automatic identification and verification of documents for airlines, airports, hotels, travel agencies and other travel partners);
- Real estate agencies (in the process of selling or renting, checking the validity of passports, documents on ownership of real estate;
- Detection of fake photos (detection of fake photos in everyday life).
Digital transformation is carried out in various sectors of the republic's economy, in particular: banking, financial, tax, and education systems. The main goal of the study is to ensure information security on various platforms, analyze their content, identify fake content and effectively use digital technologies by introducing artificial intelligence and machine learning systems on the platform.
THE MAIN PART
A.N. Romanov, E.S. Polat, V.P. Bespalko, I.G. Zakharova, E.A. Krezova, A.S. Balabina, Yu.Yu. Korolev, B. Fedorov, V.A. Gagarsky and Yu.M. Cherkasov on the development of the digital economy in the countries of the Commonwealth of Independent States, the effective use of artificial intelligence and machine learning systems.
In particular, the article by I. Klyueva discusses the practical possibilities of machine learning. An overview of industries implementing machine learning and data analysis technologies is presented. Specific use cases for machine learning are listed, from image and speech recognition to fraud detection and medical diagnostics1.
Having examined the problem of fraud in the banking sector, the authors of the scientific work come to the conclusion that it is necessary to introduce a new mechanism - artificial intelligence - in the fight against this type of crime. Having examined the risks of using AI, the authors state that this tool will significantly increase the effectiveness of combating fraud if effective measures are in place to combat cybercrime2.
In the article by the author E.M. Proydakov outlines a brief history of work in the field of artificial intelligence, describes the directions of AI, provides general information about the current state of research and development of AI systems, the main directions of research in the field of AI, AI systems in the field of public administration, and indicates the possibilities of using3.
In the article by S.B. Egorichev analyzes the use of artificial intelligence algorithms in modern banking business. The concept of artificial intelligence and the reasons for its use by banks are explained. The main areas of application of AI in banking have been identified. Features of using chatbots to improve customer service. The areas of application of artificial intelligence for customer relationship management are outlined. The possibilities of using artificial intelligence to prevent fraud in online banking are covered.
The application of AI algorithms in the areas of combating money laundering and managing bank credit risks is shown. Obstacles to the use of artificial intelligence in the banking business have been identified4.
The problem at hand is so serious that with the help of AI, we can achieve this by putting an end to many cases of fraud in banking, finance and other industries.
DISCUSSION
Artificial intelligence opens up many new opportunities that help modern manufacturers solve complex and previously impossible problems in various fields. For example, it can improve the performance and significantly expand the functionality of autonomous systems used for image analysis.
1 Klyueva Irina Alekseevna. Modern capabilities and examples of machine learning implementation. 2021 Pages: 12-32. Scientific electronic library.Elibrary.ru
2 Bagreeva Elena Gennadievna, Ismailov Nurlan Elman ogly, Bobyleva Liliya Mikhailovna. Artificial intelligence as a countermeasure to fraud in the banking sector. Eurasian advocacy. 2022;(2(57)):90
3E.M. Proydakov. The current state of artificial intelligence. DOI: 10.31249/scis/2018.00.09
4S.B.Egorycheva. Application of artificial intelligence in banking business. DOI: 10.31249/scis/2018.00.09
Artificial intelligence and machine learning are closely related to the creation of voice assistants, various robots, pictures and music. Although it has yet to achieve general intelligence matching human intelligence, many individual tasks are already being successfully performed by artificial intelligence, including neural networks and deep learning algorithms. One obvious example is image analysis: detecting objects, poses and gestures, predicting trajectory, calculating distance, etc [3].
Previously, a person could only buy a fake ID from the black market for a lot of money, but now, with the rise of e-commerce, various websites are offering their document forgery services for a fee. The lower the price of a fake document, the worse its quality. Expensive ID cards are so complex that verifying their legitimacy and preventing fraud is nearly impossible.
The ease with which fake identities can be obtained compromises the security of many services with automated identity verification systems, as well as the safety of service owners and users. There are ways to prevent fake ID fraud in machine learning. For example, you may have uploaded a fake document. If the object verification system has machine learning, the photo will be scanned by a pre-trained neural network. The fraud detection system then looks for patterns of fake documents that it has seen in many fake documents, classifies the document as fake or suspicious, and requires further verification if necessary.
Detection of counterfeit documents is primarily related to image processing. Certain techniques are used to understand the visual information of an image. CNN models are typically trained for this task, and neural networks are designed to minimize losses. CNN mimics the human visual cortex, the part of the brain responsible for processing visual information. Just as supervised learning requires a combined set of fake and real document images, the dataset must contain a sufficient number of photographs from both classes [4].
Tuning a neural network for best performance involves testing different architectures with different number of layers and filter sizes in convolutional layers. A typical convolutional architecture has four convolutional layers (Figure 1). This method has an accuracy of approximately 98% for detecting ink inconsistencies in documents forged with blue ink and 88% for black ink.
This counterfeit detection method is based on HSI, which stands for Hyperspectral Image Analysis. This method involves creating an electromagnetic spectrum map to obtain the spectrum for each pixel in the image. Another approach may be to learn and use pre-trained models, such as ImageNet, ResNet50 or VGG19 data sets. plami-based VGG16 network.
Input Conv Pool Conv Pool FC Output
Figure 1. Manuscript mismatch detection in fake documents based on convolutional
architecture 5
5https://www.v7labs.com/blog/ocr-guide
Machine learning and artificial intelligence are used to verify documents in a variety of ways. Traditionally, document verification had to be done manually, with each application reviewed by trained security professionals. This method is, of course, very slow, inaccurate and expensive.
No matter how good the security system is, processing thousands of tickets a day can never be a quick or convenient process.
People are people, and even the most experienced security professionals can make mistakes. And it is worth noting that expanding manual document verification is a very expensive and irrational undertaking [5].
Machine learning and artificial intelligence technologies help fight document forgery more effectively. A well-trained machine learning identity fraud prevention algorithm can process thousands of documents per second and filter out cases that require attention. With these AI-based algorithms, you can reduce staffing costs while increasing processing speed and security (Figure2).
Figure 2. Algorithms based on artificial intelligence6
The first step for AI document verification is to integrate the software into underlying data processing systems.
If a document needs to be verified, the new client is prompted to upload their document and the received document directly through the application [6].
Fraud detection software then uses OCR (optical character recognition) algorithms to read the information in the document and identify any typographical inconsistencies that may indicate that the document has been tampered with. At the same time, the artificial intelligence system compares the document with a database of known genuine documents and checks for visible signs of forgery [7].
CONCLUCIONS
A modern solution for online document verification based on artificial intelligence is currently being developed to reduce the risk of negative consequences associated with document forgery. This system is based on the latest developments in artificial intelligence and machine learning technologies. Using the power of AI, OCR software can also help you automate data collection from scanned documents/images.
First, AI-powered OCR enables data to be digitized in editable formats that are easy to use and compatible with organizational workflows.
6https://www.v7labs.com/blog/ocr-guide
Scanning and processing documents such as invoices, receipts, and images to capture valuable information has traditionally been done manually. OCR software solutions help businesses save time and resources on data entry and manual review. Secondly, many organizations are automating document flow, going paperless and using cloud-based digital solutions that increase profits. Document fraud detection technologies have evolved at an unprecedented rate over the years with the advent of artificial intelligence and intelligent document processing [8,9].
This is also necessary because fraudsters are increasingly able to manipulate data and hide data changes using techniques such as graph processing and deep modeling.
Third, using artificial intelligence and machine learning technologies, it would be beneficial to use counterfeit document detection techniques based on ink mismatch detection techniques in counterfeit documents based on content convolutional architecture [10,11].
REFERENCES
1. T.A. Gavrilov, V.F. Khoroshevsky. Knowledge bases of intelligent systems: Textbook. St. Petersburg: Peter, 2018. P.382.
2. J. Copeland. What is artificial intelligence? //Alan Turing. net: Reference Articles for Turing.URL:http://www.alanturing.net/turing_archive/pages/Reference%20Articles/what_is_AI/Wh at%20is%20 AI09.html.
3. K.I.Alekseevna. Modern capabilities and examples of machine learning implementation. 2021 Pages: 12-32. Scientific electronic library.Elibrary.ru
4. E.G. Bagreeva, N.E. Ismailov, L.M.Bobyleva. Artificial intelligence as a countermeasure to fraud in the banking sector. Eurasian advocacy. 2022;(2(57)):90
5. E M. Proydakov. The current state of artificial intelligence. DOI: 10.31249/scis/2018.00.09
6. S.B. Egorycheva. Application of artificial intelligence in banking business. DOI: 10.31249/scis/2018.00.09
7. N.Kh. Shoakhmedova, D.M. Yusupova. Methods of determining fake content using artificial intelligence. ICFNDS '22: The 6th International Conference on Future Networks & Distributed Systems, Tashkent, Uzbekistan, December 2022 DOI: https://doi.org/10.1145/3584202.3584234
8. D.V. Bakhteev. Artificial intelligence in forensics: status and prospects for use. Criminal procedure and criminology 2018. No. 243-47 st.
9. I.Astakhov. Artificial intelligence systems. Practical course: Textbook / M.: Binom. Laboratory work, 2017. P. 292.
10. I.G. Sidorkin. Artificial intelligence systems: Textbook / M.: KnoRus, 2015. P. 249.
11. L.N. Yasnitsky. Introduction to artificial intelligence: A textbook for students. above flying establishments / M.: Publishing center "Academy", 2015. P. 176.
12. https://www.v7labs.com/blog/ocr-guide
