基本信息

姓名：高嵩

职称：讲师（硕士生导师）

系所：软件工程

研究领域：计算机视觉、人工智能安全、模型压缩

电子邮箱：gaos@ynu.edu.cn

个人简介

高嵩，2021年毕业于云南大学信息学院，获博士学位。2024年于云南大学系统科学博士后流动站出站。同年12月入职英国365集团官网，从事教学科研工作。主要研究方向是计算机视觉、人工智能安全、模型压缩等，近3年以第一或通信作者身份发表高水平期刊或会议论文10余篇，包括IEEE TDSC、IEEE TSC、IEEE TETCI、PR等。

教育经历

2017/09-2021/12，云南大学，信息与通信工程，博士，导师：姚绍文

2014/09-2017/07，云南大学，计算机应用技术，硕士，导师：周丽华

2009/09-2013/07，中国地质大学（武汉），软件工程，本科

科研与学术工作经历：

2022/01-2024/09，云南大学系统科学博士后流动站，博士后，合作导师：周维

2019/11-2020/11，悉尼科技大学，联合培养博士，导师：余水

主要研究领域

1.对抗样本生成与攻击策略：研究面向图像分类、目标检测等任务的对抗样本生成方法，探索高效、可迁移的攻击算法，揭示深度学习模型在对抗扰动下的脆弱性机制。

2.深度学习模型鲁棒性增强：致力于提升模型在复杂对抗环境下的可靠性，研究基于对抗训练、特征增强与模型架构优化的防御策略，增强模型对未知扰动的泛化能力。

3.轻量级模型鲁棒性优化：针对资源受限场景，开展模型压缩与加速过程中的鲁棒性保持研究，探索轻量化网络在效率与安全性之间的平衡机制。

4.端侧攻防框架设计与应用：面向无人机等边缘计算平台，构建适用于端侧部署的轻量级攻防框架，实现低开销、实时的对抗攻击检测与防御，推动安全AI在真实物理场景中的落地。

部分学术成果

[1]Song Gao, Ruxin Wang, Xiaoxuan Wang, Shui Yu, Yunyun Dong, Shaowen Yao and Wei Zhou, "Detecting Adversarial Examples on Deep Neural Networks With Mutual Information Neural Estimation", inIEEE Transactions on Dependable and Secure Computing, vol. 20, no. 6, pp. 5168-5181, 2023.

[2]Song Gao, Xiaoxuan Wang, Bingbing Song, Renyang Liu, Shaowen Yao, Wei Zhou and Shui Yu, “Exploiting Type I Adversarial Examples to Hide Data Information: A New Privacy-Preserving Approach”, inIEEE Transactions on Emerging Topics in Computational Intelligence, doi: 10.1109/TETCI.2024.33678 12, 2024.

[3]Song Gao, Shui Yu, Liwen Wu, Shaowen Yao and Xiaowei Zhou, "Detecting adversarial examples by additional evidence from noise domain", inIET Image Processing, vol. 16, pp. 378-392, 2022.

[4]Song Gao, Shaowen Yao and Ruidong Li, “Transferable Adversarial Defense by Fusing Reconstruction Learning and Denoising Learning” inProceedings of the IEEE Conference on Computer Communications Workshops, 2021: 1-6.

[5]Song Gao, Shui Yu, Shaowen Yao, “An efficient protein homology detection approach based on seq2seq model and ranking”, inBiotechnology&Biotechnological Equipment, 25(1), 2021: 663-640.

[6]Jinming Cui,Song Gao*, Tao Lv, Jun Ji, Shaowen Yao, Wei Zhou. “Dual-label guided unrestricted target attack with diffusion model”, inNeurocomputing, 665, 2026.

[7]Jun Ji,Song Gao*and Wei Zhou, “Transferable adversarial sample purification by expanding the purification space of diffusion models”, inThe Visual Computer, doi: 10.1007/s00371-023-03253-7, 2024.

[8]Weiwei Zeng,Song Gao*, Wei Zhou, Yunyun Dong and Ruxin Wang, “Improving the Adversarial Robustness of Object Detection with Contrastive Learning”, inProceedings of the Chinese Conference on Pattern Recognition and Computer Vision, 2023, pp. 29-40.

[9]Liwen Wu,Song Gao, Shaowen Yao ; Feng Wu, Jie Li; Yunyun Dong and Yunqi Zhang, “Gm-PLoc: A Subcellular Localization Model of Multi-Label Protein Based on GAN and DeepFM”, inFrontiers in Genetics, vol. 13, 2022.

[10]Fuyi Hu, Jin Zhang,Song Gao, Yu Lin, Wei Zhou and Ruxin Wang, “An efficient training-from-scratch framework with BN-based structural”, inPattern Recognition, 153, 2024: 110546.

[11]Bingbing Song, Ruxin Wang,Song Gao, Yunyun Dong, Ling Liu and Wei Zhou, “Securing Deep Learning as a Service Against Adaptive High Frequency Attacks With MMCAT”, inIEEE Transactions on Sevices Computing, vol. 16, no. 5, pp. 3723-3735, 2023.

[12]Yi Zhao, Xin Jin,Song Gao, Liwen Wu, Shaowen Yao and Qian Jiang, “TAN-GFD: Generalizing Face Forgery Detection based on Texture Information and Adaptive Noise Mining” inApplied Intelligence, vol. 53, no. 16, pp.19007-19027,2023.

[13]Haoran Li, Jinhong Zhang,Song Gao, Liwen Wu, Wei Zhou and Ruxin Wang, “Towards Query-limited Adversarial Attacks on Graph Neural Networks”, inProceedings of the International Conference on Tools With Artificial Intelligence, 2022, pp. 516-521.

[14]Feiyang Qin, Wenqi Na,Song Gaoand Shaowen Yao, “Sigma-UAP: An Invisible Semi-Universal Adversarial Attack Against Deep Neural Networks”, inProceedings of the International Conference on Pattern Recognition, Machine Vision and Intelligent Algorithms, 2023, pp. 34-41.

[15]Yu Shi, Tan Hu,Song Gao, Yunyun Dong, Wei Zhou and Ruxin Wang, “MMID: Combining Maximized the Mutual Information and Diffusion Model for Image Super-Resolution”, inProceedings of the Asian Conference on Pattern Recognition, 2023, pp. 381-395.

部分专利

[1]高嵩,吕涛,崔金明,杨家勋,姚绍文, “多图神经网络集成的图相似度评估方法”,发明专利,中国, 2026/01/09.

[2]高嵩,杨家勋,周彦锦,吕涛,吴峰,姚绍文, “基于梯度分解的机器遗忘方法”,发明专利,中国, 2026/01/09.

[3]高嵩,曾威威,铁清元,王汝欣,石宇, “基于对比学习的目标检测模型对抗训练方法”,发明专利,中国, 2025/06/10.

[4]高嵩,周维,王晓璇,董云云,武丽雯,姚绍文, “攻击无依赖的可迁移对抗样本检测方法”,发明专利,中国, 2024/05/14.

主要科研项目

[1]面向目标检测系统的对抗样本防御技术研究, “彩云博士后计划”创新项目, 5万元,已结项,主持.

[2]零知识依赖的对抗样本泛化防御理论及关键技术研究,云南省科技厅基础研究专项青年项目, 5万元,在研,主持.

[3]大数据环境下多模态强化学习策略搜索理论与方法研究,国家自然科学基金委员会地区科学基金项目, 2022-01-01至2025-12-31, 37万元,已结项,参与.

[4]基于ICM脉冲时间编码的蛋白质序列特征提取及二级结构预测研究,国家自然科学基金委员会地区科学基金项目, 2019-01-01至2022-12-31, 40万元,已结项,参与.

[5]结合多模态图像融合的低空无人机检测技术研究,国家自然科学基金委员会地区科学基金项目, 2023-01-01至2026-12-31, 33万元,在研,参与.