AI Safety in Practice: Enhancing Adversarial Robustness in Multimodal Image Captioning
This work addresses safety concerns for deploying multimodal AI in critical applications, but it is incremental as it builds on existing adversarial training techniques.
The paper tackled the vulnerability of multimodal image captioning models to adversarial attacks by using FGSM-generated adversarial examples and adversarial training, achieving improved robustness on Flickr8k and COCO datasets with a selective training method that matches full training performance while being more efficient.
Multimodal machine learning models that combine visual and textual data are increasingly being deployed in critical applications, raising significant safety and security concerns due to their vulnerability to adversarial attacks. This paper presents an effective strategy to enhance the robustness of multimodal image captioning models against such attacks. By leveraging the Fast Gradient Sign Method (FGSM) to generate adversarial examples and incorporating adversarial training techniques, we demonstrate improved model robustness on two benchmark datasets: Flickr8k and COCO. Our findings indicate that selectively training only the text decoder of the multimodal architecture shows performance comparable to full adversarial training while offering increased computational efficiency. This targeted approach suggests a balance between robustness and training costs, facilitating the ethical deployment of multimodal AI systems across various domains.