Adversarial hujum usullari yordamida obyektlarni aniqlash modellarini chalg‘itish va ularning barqarorligini baholash

Mualliflar

Kalit so‘zlar: FGSM, YOLOv5, PGD, Faster R-CNN, ResNet18, CNN

Annotatsiya

Ushbu tadqiqotda konvolyutsion neyron tarmoqlarga asoslangan obyekt aniqlash tizimlari, xususan, YOLOv5 modeliga qarshi adversarial hujumlar samaradorligi o‘rganildi. Hujum usullari sifatida Fast Gradient Sign Method (FGSM) hamda uning takomillashtirilgan variantlari (Iterative FGSM va PGD) qo‘llanildi. ResNet18 modeli yordamida hosil qilingan adversarial perturbatsiyalar asosida original rasm va hujumga uchragan rasmning obyekt aniqlash natijalari solishtirildi. Natijalar shuni ko‘rsatdiki, hatto kichik miqdordagi shovqin ham obyekt aniqlash modelini sezilarli darajada chalg‘itishi mumkin. Adversarial hujumlar sun’iy intellekt modellarining, ayniqsa chuqur o‘rganishga asoslangan tasvirni tanish tizimlarining asosiy zaifliklaridan biri hisoblanadi. FGSM va PGD kabi hujumlar model kirishini kichik perturbatsiyalar orqali o‘zgartirib, modelni noto‘g‘ri klassifikatsiya qilishga majbur qiladi.

Foydalanilgan adabiyotlar

1. Goodfellow I., Shlens J., Szegedy C. Explaining and harnessing adversarial examples. // arXiv preprint arXiv:1412.6572. 2014. https://arxiv.org/abs/1412.6572

2. Szegedy C., Zaremba, W., Sutskever, I., Bruna, J., Erhan, D., Goodfellow, I., & Fergus, R. Intriguing properties of neural networks. // arXiv preprint arXiv:1312.6199. 2013. https://arxiv.org/abs/1312.6199

3. Kurakin A., Goodfellow I., Bengio, S. Adversarial machine learning at scale. // arXiv preprint arXiv:1611.01236. 2016. https://arxiv.org/abs/1611.01236

4. Madry A., Makelov A., Schmidt L., Tsipras D., Vladu A. Towards deep learning models resistant to adversarial at- tacks. Proceedings of the 6th International Conference on Learning Representations (ICLR). 2018.

5. Papernot N., McDaniel P., Goodfellow I., Jha S., Celik Z.B., Swami A. Practical black-box attacks against deep learning systems using adversarial examples. // IEEE Symposium on Security and Privacy (SP), 2017. – Р. 1-18.