IASTEM INTERNATIONAL CONFERENCE
Medina,Saudi Arabia 14-09-2026
1 paper
-
Preventing Cyber Risks for Children Using Machine Learning (E.C.B)
Batool Albadan, Dr. Syed Mujahid
Paper ID IA-ICRAETMEDI-140926-17643 · Pages 45-49
Download PDF: Preventing Cyber Risks for Children Using Machine Learning (E.C.B)Abstract
With the growing use of mobile devices among children, ensuring safe access to age-appropriate applications has become essential. Traditional parental controls and age-gating mechanisms are often either too restrictive or easily bypassed. This study explores an innovative approach to child protection in digital spaces by implementing image-based age detection, using facial recognition and classification to estimate a child’s age when they attempt to access certain applications. One of the primary challenges in this approach lies in balancing the misclassification costs associated with age estimation errors, which can result in underestimating or overestimating a child’s age. Each misclassification has distinct consequences: underestimation could limit children’s access to suitable content, while overestimation could expose them to inappropriate material. To address these challenges, this research utilizes a dual-model framework, combining a global model pre-trained on extensive datasets with a localized model on the user’s device, tailored for the specific user environment. The global model benefits from broader generalization, while the local model adapts to individual conditions, allowing for enhanced accuracy in age estimation. Privacy-preserving techniques, such as data encryption and anonymization, ensure that user data is protected, with parents given the option to control data sharing preferences. This study’s findings suggest that the methodology effectively balances age classification accuracy, privacy, and parental control, offering a scalable solution to protect children in digital environments. Keywords - Image Classification, Misclassification Cost, Machine Learning Models, Cost-Sensitive Framework, Privacy Protection, Misclassification Errors, Privacy-Preserving Techniques, Cost-Sensitive Learning, Digital Content Access, Child Protection Systems, Deep Learning