Backdoor attacks on deep neural networks are commonly evaluated by attack success rate and benign accuracy, yet their stealthiness remains inconsistently defined and measured. Existing work often studies only one observable aspect of stealthiness, such as trigger imperceptibility, feature-space separability, parameter-space concentration, or evasion of particular defenses. This creates a fragmented picture: an attack may appear stealthy under one metric while leaving conspicuous footprints elsewhere. This paper presents a tri-space measurement study of backdoor stealthiness in image classification. We distinguish latent stealthiness---the ideal but unobservable difficulty of distinguishing a backdoored system from a benign one---from measurable footprints in three spaces: input, feature, and parameter space. Under this framework, we evaluate representative backdoor attacks across four datasets and three model architectures using a broad suite of footprint metrics.
Our results show that stealthiness is not monolithic: attacks that are quiet in one space are often loud in another, and existing attacks rarely achieve consistently small footprints across all three spaces. We further analyze agreement and disagreement among stealthiness metrics and study how footprint measurements interact with attack success rate. Based on observed blind spots in existing measurements, we introduce two diagnostic metrics: Class-Specific Davies--Bouldin Index (CDBI), which captures class-conditional feature overlap, and TAC--UCLC Product (TUP), which identifies channels that are both trigger-responsive and downstream influential. These metrics are intended to augment, not replace, the broader measurement suite. Finally, we provide practical guidance for future stealthiness evaluations.
Towards Backdoor Stealthiness in Model Parameter Space
Radboud University(拉德博德大学);Delft University of Technology(代尔夫特理工大学);Radboud University, Nijmegen, The Netherlands(拉德博德大学);University of Zagreb Faculty of Electrical Engineering and Computing(Zagreb 工程与计算学院)
College of Control Science and Engineering, China University of Petroleum (East China)(控制科学与工程学院,中国石油大学(华东));College of Computer Science and Electronic Engineering, Hunan University(计算机科学与电子工程学院,湖南大学)
SNEAKDOOR: Stealthy Backdoor Attacks against Distribution Matching-based Dataset Condensation
School of Computer Science and Technology, Xi’an Jiaotong University(西安交通大学计算机科学与技术学院);National Key Laboratory of Human-Machine Hybrid Augmented Intelligence, Xi’an Jiaotong University(西安交通大学人机混合增强智能全国重点实验室)
Backdoor Attack Against Vision Transformers via Attention Gradient-Based Image Erosion
University of Electronic Science and Technology of China(电子科技大学);Laboratory Of Intelligent Collaborative Computing, University of Electronic Science and Technology of China(电子科技大学智能协同计算实验室);School of Computer Science and Engineering, University of Electronic Science and Technology of China(电子科技大学计算机科学与工程学院)
Transferable Backdoor Attacks for Code Models via Sharpness-Aware Adversarial Perturbation
Nanjing University of Posts and Telecommunications(南京邮电大学);State Key Laboratory of Tibetan Intelligence(西藏智能重点实验室);Jiangsu Provincial Key Laboratory of Internet of Things Intelligent Perception and Computing(江苏省物联网智能感知与计算重点实验室);Anhui Province Key Laboratory of Cyberspace Security Situation Awareness and Evaluation(安徽省网络安全态势感知与评估重点实验室);University of Technology Sydney(悉尼科技大学);The University of Melbourne(墨尔本大学);Griffith University(格里菲斯大学)
A Spatiotemporal Stealthy Backdoor Attack against Cooperative Multi-Agent Deep Reinforcement Learning
School of Cybersecurity, Northwestern Polytechnical University(西北工业大学网络空间安全学院);Research & Development Institute of Northwestern Polytechnical University in Shenzhen(西北工业大学深圳研究院)
Does Few-shot Learning Suffer from Backdoor Attacks?
Institute of Information Engineering, CAS(中国科学院信息工程研究所);University of Chinese Academy of Sciences(中国科学院大学);Nanyang Technological University(南洋理工大学);University of Oxford(牛津大学);National University of Singapore(新加坡国立大学);Sun Yat-sen University(中山大学)
MirageBackdoor: A Stealthy Attack that Induces Think-Well-Answer-Wrong Reasoning
Institute of Information Engineering, Chinese Academy of Sciences(中国科学院信息工程研究所);School of Cyber Security, University of Chinese Academy of Sciences(中国科学院大学网络空间安全学院);Beijing University of Posts and Telecommunications(北京邮电大学)
Towards Stealthy and Effective Backdoor Attacks on Lane Detection: A Naturalistic Data Poisoning Approach
State Key Laboratory of Intelligent Vehicle Safety Technology, Changan Automobile(长安汽车智能车辆安全技术国家重点实验室);National University of Singapore (Chongqing) Research Institute, Chongqing(新加坡国立大学(重庆)研究院);Hong Kong University of Science and Technology (Guangzhou), Guangzhou(香港科技大学(广州));National University of Singapore, Singapore(新加坡国立大学);Ningbo University, Ningbo(宁波大学);Sun Yat-Sen University, Guangzhou(中山大学);Chongqing Changan Automobile Co., Ltd., Chongqing(重庆长安汽车有限公司)
LaserGuider: A Laser Based Physical Backdoor Attack against Deep Neural Networks
ShanghaiTech University(上海科技大学);Pengcheng Laboratory(鹏城实验室);Key Laboratory of System Software (Chinese Academy of Sciences) and State Key Laboratory of Computer Science, Institute of Software, Chinese Academy of Science(中国科学院软件研究所系统软件重点实验室与计算机科学国家重点实验室);Nanjing Institute of Software Technology(南京软件技术研究院)
University of Science and Technology of China(中国科学技术大学);Microsoft Research(微软研究院);Beijing Electronic Science and Technology Institute(北京电子科技学院);Wormpex AI Research LLC(Wormpex人工智能研究有限责任公司)
Shortcuts Arising from Contrast: Effective and Covert Clean-Label Attacks in Prompt-Based Learning
School of Electronic Engineering(电子工程学院);Beijing University of Posts and Telecommunications(北京邮电大学);Centre for Frontier AI Research (CFAR)(前沿人工智能研究中心);Institute of High Performance Computing (IHPC)(高性能计算研究所);Agency for Science Technology and Research (A*STAR)(新加坡科技研究局);School of Information and Communication(信息与通信学院)
Semantic-level Backdoor Attack against Text-to-Image Diffusion Models
College of Computer Science and Artificial Intelligence, Fudan University, Shanghai, China(复旦大学计算机科学与人工智能学院,上海,中国);State Key Laboratory of Integrated Services Networks, Xidian University, Xian, China(集成服务网络国家重点实验室,西安电子科技大学,西安,中国);University of Electronic Science and Technology of China, Sichuan, China(电子科技大学,四川,中国);Ajou University, Gyeonggi-do, South Korea(庆尚道,韩国,全州大学)
BadCM: Invisible Backdoor Attack Against Cross-Modal Learning
School of Computer Science and Technology, Harbin Institute of Technology, Shenzhen(哈尔滨工业大学(深圳)计算机科学与技术学院);Peng Cheng Laboratory(鹏城实验室);School of Electronic and Information Engineering, Tongji University(同济大学电子与信息工程学院);School of Computer Science and Engineering, University of Electronic Science and Technology of China(电子科技大学计算机科学与工程学院)