A joint research team led by Professor Oh Hyun-ok of the Department of Information Systems at Hanyang University and Professor Kim Ji-hye of the School of Electronic Engineering at Kookmin University received the Best Paper Award from the prestigious international journal IEEE Transactions on Dependable and Secure Computing (TDSC) for their verifiable artificial intelligence (Verifiable 토토 사이트 바카라) technology, vCNN. The study was recognized for its breakthrough in proving the correctness of 토토 사이트 바카라 inference results without disclosing input data or the underlying model.

As AI technologies become increasingly applied to fields such as medical diagnostics, financial fraud detection, and public services, there is a growing need to ensure inference accuracy while protecting sensitive data (e.g., medical images, transaction histories) and maintaining confidentiality of model weights, which are valuable intellectual property. Existing approaches either required disclosing personal or proprietary information for validation or involved cryptographic techniques that demanded excessive computation time—hindering their scalability.

To address these challenges, the research team developed vCNN (Verifiable Convolutional Neural Network), which leverages zk-SNARKs (zero-knowledge succinct non-interactive arguments of knowledge) to prove that an AI model has performed inference accurately—without revealing the input or model. Notably, the team designed a novel proof method for convolution operations, the core computation in CNNs, reducing the time complexity from O(l·n) (where l is kernel size and n is data size) to O(l + n). As a result, vCNN achieved a 20-fold speed increase for the MNIST model and an 18,000-fold increase for the VGG16 model, all while m토토 사이트 바카라nt토토 사이트 바카라ning mathematically proven security.

Professor Oh Hyun-ok of Hanyang University, who led the study, remarked, “vCNN opens a new chapter for trustworthy AI. We aim to further develop this technology to deliver reliable AI services in real-world applications such as healthcare and finance.”

TDSC, published by the IEEE Computer Society, is a top-tier (Q1) journal in the field of dependability and security. Each year, it selects only one paper among all published that stands out in originality and impact to receive the Best Paper Award. This year’s honor was awarded solely to the vCNN paper.

The research was supported by the Ministry of Science and ICT and the Institute of Information & Communications Technology Planning & Evaluation (IITP). The paper, titled “vCNN: Verifiable Convolutional Neural Network Based on zk-SNARKs”, lists Dr. Lee Seung-hwa (Kookmin University) as the first author, Dr. Ko Han-kyung (Hanyang University) as a co-author, and Professors Oh Hyun-ok (HYU) and Kim Ji-hye (Kookmin University) as co-corresponding authors.