Siber güvenlik tehditlerini ve fidye yazılım saldırılarını makine öğrenimiyle önlemeye yönelik yeni bir çalışma

The purpose of this master's research thesis is to categorize the cutting-edge AI-based cyber security network to reduce cybercrimes globally and provide a resilient cyberattack-free environment for consumers. One of the most often used real applications in areas like cost control, traffic scene analysis, and thought attack identification is attacks restriction. Contextual information, which in our study is described as the relationship between the cyber security and the Rivest-Shamir-Adleman, is important to get total comprehension along with the information on the owner vehicle. The research study suggested a KNN classifier for continuously limiting the position of organizations and digital protection in the area. The modified districts are passed to the Rivest-Shamir-Adleman after being preprocessed to reduce commotion (RSA). The preprocessing of the RSA separates associated components, divides them into lines of attack, and channels them by size. The well-known technique used in the research is RSA. The user is advised to acquire a new attack of that portion of the assault if the RSA reported a low confidence score for at least one attack region. The new attack is then compared to the existing attack, and the attack processing steps are once more carried out. If all locations have a sufficiently high score or enough retries have been used up, the security cycle ends. The outcome includes the type of assault that was discovered, a list of the districts that were subject to the removed assault, and the accuracy of the RSA calculation through KNN, which was calculated across 300 preparation and testing iterations. The proposed framework was created using 75% of the available data, 20% of which was used for testing, and 5% for validation. We used a Cyber Data Dataset made up of multiple classes that addressed a portion of all current forms of digital protection to illustrate the meaning of the proposed component vector. Multiple machine learning toolboxes were utilized for the research effort, which was implemented and carried out in the MATLAB programming language. Our strategy has three practical head networks, so a potential preparation risk is prompted by the start to finish and synchronous preparation. It is incredibly challenging to determine which head configuration is producing a problem on the off chance that the model doesn't come together. As a result, in our plan cycle, we gradually added each helpful head to make sure the single head plan idea was functioning correctly before extending the model with the remaining useful heads.