Preprints

1. Ma, J., Aldeen, M., Salas, C., Luo, F., Chowdhury, M., Pese, M., & Cheng, L. (2025). DisPatch: Disarming Adversarial Patches in Object Detection with Diffusion Models. arXiv:2509.04597. https://arxiv.org/pdf/2509.04597

   PDF arXiv BIB Abstract

   @unpublished{ma2025dispatch,
     title = {DisPatch: Disarming Adversarial Patches in Object Detection with Diffusion Models},
     author = {Ma, Jin and Aldeen, Mohammed and Salas, Christopher and Luo, Feng and Chowdhury, Mashrur and Pese, Mert and Cheng, Long},
     note = {arXiv:2509.04597},
     doi = {2509.04597},
     url = {https://arxiv.org/pdf/2509.04597},
     year = {2025}
   }
   

   We introduce DisPatch, the first diffusion-based defense framework for object detection against adversarial patch attacks.

Refereed Conference Proceedings

1. Han, X., Ma, J., Zhang, J., Liu, K., & Luo, F. (2025). Understanding the Constraints of RAG-Based Medical LVLMs: A Case Study in Ophthalmic Report Generation. 2025 IEEE International Conference on Data Mining Workshops (ICDMW), 96–102.

   DOI BIB Abstract

   @inproceedings{han2025understanding,
     title = {Understanding the Constraints of RAG-Based Medical LVLMs: A Case Study in Ophthalmic Report Generation},
     author = {Han, Xiaoyan and Ma, Jin and Zhang, Jinghan and Liu, Kunpeng and Luo, Feng},
     booktitle = {2025 IEEE International Conference on Data Mining Workshops (ICDMW)},
     pages = {96--102},
     year = {2025},
     doi = {10.1109/ICDMW69685.2025.00017}
   }
   

   Medical Large Vision-Language Models (Med-LVLMs) offer substantial potential to advance disease diagnosis by integrating visual medical data with textual clinical knowledge. However, they still face significant challenges with factual hallucination because their knowledge is limited to what was encoded during training. While Retrieval-Augmented Generation (RAG) has emerged as a promising solution, its effectiveness is highly dependent on the quality of retrieved documents, where irrelevant or noisy information can compromise the reliability of generated responses.

2. Yan, J., Liao, S., Ma, J., Aldeen, M., Kumar, S., & Cheng, L. (2025). No Way to Sign Out? Unpacking Non-Compliance with Google Play’s App Account Deletion Requirements. 34th USENIX Security Symposium (USENIX Security 25), 3277–3296. https://www.usenix.org/system/files/usenixsecurity25-yan-jingwen.pdf

   PDF BIB Abstract

   @inproceedings{yan2025no,
     title = {No Way to Sign Out? Unpacking Non-Compliance with Google Play's App Account Deletion Requirements},
     author = {Yan, Jingwen and Liao, Song and Ma, Jin and Aldeen, Mohammed and Kumar, Salish and Cheng, Long},
     booktitle = {34th USENIX Security Symposium (USENIX Security 25)},
     pages = {3277--3296},
     year = {2025},
     url = {https://www.usenix.org/system/files/usenixsecurity25-yan-jingwen.pdf}
   }
   

   We conducted the first study investigating non-compliance issues with Google Play’s app account deletion requirements.

3. Ma, J., Aldeen, M. S., Luo, F., & Cheng, L. (2025). Few-Shot Detection of Hate Videos Using Multi-Modal Large Language Models. Proceedings of the 1st ACM Workshop on Deepfake, Deception, and Disinformation Security, 32–35. https://dl.acm.org/doi/pdf/10.1145/3733813.3764370

   PDF DOI BIB Abstract

   @inproceedings{ma2025few,
     title = {Few-Shot Detection of Hate Videos Using Multi-Modal Large Language Models},
     author = {Ma, Jin and Aldeen, Mohammed Shujaa and Luo, Feng and Cheng, Long},
     booktitle = {Proceedings of the 1st ACM Workshop on Deepfake, Deception, and Disinformation Security},
     pages = {32--35},
     year = {2025},
     doi = {10.1145/3733813.3764370},
     url = {https://dl.acm.org/doi/pdf/10.1145/3733813.3764370}
   }
   

   We present a method for hate video detection using multi-modal large language models (MLLMs), without time-consuming training.

4. Ma, J., Enan, A., Cheng, L., & Chowdhury, M. (2025). Understanding the Risks of Asphalt Art on the Reliability of Surveillance Perception Systems. Transportation Research Board (TRB) Annual Meeting. https://arxiv.org/pdf/2508.02530

   PDF BIB Abstract

   @inproceedings{ma2025understanding,
     title = {Understanding the Risks of Asphalt Art on the Reliability of Surveillance Perception Systems},
     author = {Ma, Jin and Enan, Abyad and Cheng, Long and Chowdhury, Mashrur},
     booktitle = {Transportation Research Board (TRB) Annual Meeting},
     year = {2025},
     url = {https://arxiv.org/pdf/2508.02530}
   }
   

   We investigate the impact of asphalt art on the reliability of surveillance perception systems.

5. Ma, J., Pang, S., Yang, B., Zhu, J., & Li, Y. (2020). Spatial-content image search in complex scenes. Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2503–2511. https://openaccess.thecvf.com/content_WACV_2020/papers/Ma_Spatial-Content_Image_Search_in_Complex_Scenes_WACV_2020_paper.pdf

   PDF BIB Abstract

   @inproceedings{ma2020spatial,
     title = {Spatial-content image search in complex scenes},
     author = {Ma, Jin and Pang, Shanmin and Yang, Bo and Zhu, Jihua and Li, Yaochen},
     booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
     pages = {2503--2511},
     year = {2020},
     url = {https://openaccess.thecvf.com/content_WACV_2020/papers/Ma_Spatial-Content_Image_Search_in_Complex_Scenes_WACV_2020_paper.pdf}
   }
   

   We develop a novel visually similar spatial-semantic method, namely spatial-content image search.