mlsanityData to 14 Oct 2026

Poison-splat: Computation Cost Attack on 3D Gaussian Splatting

arXiv 2410.08190 ↗First posted 10 Oct 2024cs.CV cs.CR cs.GRcited by 18

Not on the map. Why?

Abstract (arXiv)

3D Gaussian splatting (3DGS), known for its groundbreaking performance and efficiency, has become a dominant 3D representation and brought progress to many 3D vision tasks. However, in this work, we reveal a significant security vulnerability that has been largely overlooked in 3DGS: the computation cost of training 3DGS could be maliciously tampered by poisoning the input data. By developing an attack named Poison-splat, we reveal a novel attack surface where the adversary can poison the input images to drastically increase the computation memory and time needed for 3DGS training, pushing the algorithm towards its worst computation complexity. In extreme cases, the attack can even consume all allocable memory, leading to a Denial-of-Service (DoS) that disrupts servers, resulting in practical damages to real-world 3DGS service vendors. Such a computation cost attack is achieved by addressing a bi-level optimization problem through three tailored strategies: attack objective approximation, proxy model rendering, and optional constrained optimization. These strategies not only ensure the effectiveness of our attack but also make it difficult to defend with simple defensive measures. We hope the revelation of this novel attack surface can spark attention to this crucial yet overlooked vulnerability of 3DGS systems. Our code is available at https://github.com/jiahaolu97/poison-splat .

Key facts. Cited by 18 papers in the map (data through 14 Oct 2026). Builds on 65 of its 91 references that are in the map. First posted on arXiv 10 Oct 2024.

Builds on

65 of 91 references.

References · 65

  1. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,415
  2. 20233D Gaussian Splatting for Real-Time Radiance Field Renderingcited by 6,226
  3. 2013Intriguing properties of neural networkscited by 5,155
  4. 2016Membership Inference Attacks against Machine Learning Modelscited by 2,281
  5. 2020Extracting Training Data from Large Language Modelscited by 2,189
  6. 2021Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fieldscited by 2,032
  7. 2017Understanding Black-box Predictions via Influence Functionscited by 1,741
  8. 20234D Gaussian Splatting for Real-Time Dynamic Scene Renderingcited by 1,157
  9. 2023DreamGaussian: Generative Gaussian Splatting for Efficient 3D Content Creationcited by 843
  10. 2023Dynamic 3D Gaussians: Tracking by Persistent Dynamic View Synthesiscited by 737
  11. 2023Extracting Training Data from Diffusion Modelscited by 732
  12. 2020Inverting Gradients -- How easy is it to break privacy in federated learning?cited by 712
38 more
  1. 2023Scaffold-GS: Structured 3D Gaussians for View-Adaptive Renderingcited by 695
  2. 2017Evasion Attacks against Machine Learning at Test Timecited by 689
  3. 2023Mip-Splatting: Alias-free 3D Gaussian Splattingcited by 667
  4. 2016Stealing Machine Learning Models via Prediction APIscited by 619
  5. 2018ML-Leaks: Model and Data Independent Membership Inference Attacks and Defenses on Machine Learning Modelscited by 498
  6. 2023Real-time Photorealistic Dynamic Scene Representation and Rendering with 4D Gaussian Splattingcited by 476
  7. 2019Evading Defenses to Transferable Adversarial Examples by Translation-Invariant Attackscited by 442
  8. 2021Deep Long-Tailed Learning: A Surveycited by 417
  9. 2023Gaussian Splatting SLAMcited by 415
  10. 2023Gaussian Grouping: Segment and Edit Anything in 3D Scenescited by 393
  11. 2020Label-Only Membership Inference Attackscited by 388
  12. 2023SplaTAM: Splat, Track & Map 3D Gaussians for Dense RGB-D SLAMcited by 388
  13. 2023PhysGaussian: Physics-Integrated 3D Gaussians for Generative Dynamicscited by 377
  14. 2023Text-to-3D using Gaussian Splattingcited by 299
  15. 2023GS-SLAM: Dense Visual SLAM with 3D Gaussian Splattingcited by 278
  16. 2018Adversarial Attacks and Defences: A Surveycited by 259
  17. 2023GaussianDreamer: Fast Generation from Text to 3D Gaussians by Bridging 2D and 3D Diffusion Modelscited by 255
  18. 2018Manipulating Machine Learning: Poisoning Attacks and Countermeasures for Regression Learningcited by 225
  19. 2024A Survey on 3D Gaussian Splattingcited by 224
  20. 2024Mini-Splatting: Representing Scenes with a Constrained Number of Gaussianscited by 213
  21. 2024GaussianPro: 3D Gaussian Splatting with Progressive Propagationcited by 200
  22. 2017Towards Poisoning of Deep Learning Algorithms with Back-gradient Optimizationcited by 197
  23. 2020A Survey of Privacy Attacks in Machine Learningcited by 146
  24. 2023Multi-Scale 3D Gaussian Splatting for Anti-Aliased Renderingcited by 138
  25. 2018Stronger Data Poisoning Attacks Break Data Sanitization Defensescited by 135
  26. 2024Recent Advances in 3D Gaussian Splattingcited by 131
  27. 2024Revising Densification in Gaussian Splattingcited by 128
  28. 2023GART: Gaussian Articulated Template Modelscited by 121
  29. 2020MetaPoison: Practical General-purpose Clean-label Data Poisoningcited by 120
  30. 2023Drivable 3D Gaussian Avatarscited by 120
  31. 2021Adversarial Examples Make Strong Poisonscited by 119
  32. 2020Sponge Examples: Energy-Latency Attacks on Neural Networkscited by 111
  33. 2018Exploring Connections Between Active Learning and Model Extractioncited by 85
  34. 2016Data Poisoning Attacks on Factorization-Based Collaborative Filteringcited by 83
  35. 2024AbsGS: Recovering Fine Details for 3D Gaussian Splattingcited by 82
  36. 2022I Know What You Trained Last Summer: A Survey on Stealing Machine Learning Models and Defencescited by 75
  37. 2023Bag of Tricks for Training Data Extraction from Language Modelscited by 70
  38. 2024Inducing High Energy-Latency of Large Vision-Language Models with Verbose Imagescited by 69

50 most cited of 65 All 65 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2024: 2; 2025: 5; 2026: 11.

18 papers.

Sort:

By area:

  1. 2024MEGA: Memory-Efficient 4D Gaussian Splatting for Dynamic Scenescited by 47
  2. 2025Matrix-3D: Omnidirectional Explorable 3D World Generationcited by 35
  3. 2025RemedyGS: Defend 3D Gaussian Splatting against Computation Cost Attackscited by 7
  4. 2026Safety in Embodied AI: A Survey of Risks, Attacks, and Defensescited by 7
  5. 2024Faster and Better 3D Splatting via Group Trainingcited by 6
  6. 2025StealthAttack: Robust 3D Gaussian Splatting Poisoning via Density-Guided Illusionscited by 6
  7. 2026Neural Gabor Splatting: Enhanced Gaussian Splatting with Neural Gabor for High-frequency Surface Reconstructioncited by 2
  8. 2026Spectral Defense Against Resource-Targeting Attack in 3D Gaussian Splattingcited by 1
  9. 2026PoInit-of-View: Poisoning Initialization of Views Transfers Across Multiple 3D Reconstruction Systemscited by 1
  10. 20253D Gaussian Splat Vulnerabilitiescited by 0
  11. 2025ComplicitSplat: Downstream Models are Vulnerable to Blackbox Attacks by 3D Gaussian Splat Camouflagescited by 0
  12. 2026DefenseSplat: Enhancing the Robustness of 3D Gaussian Splatting via Frequency-Aware Filteringcited by 0
6 more
  1. 2026AdvSplat: Adversarial Attacks on Feed-Forward Gaussian Splatting Modelscited by 0
  2. 2026PatchPoison: Poisoning Multi-View Datasets to Degrade 3D Reconstructioncited by 0
  3. 2026Optimizing 3D Gaussian Splatting via Point Cloud Upsamplingcited by 0
  4. 2026Characterizing Detectability in 3DGS Poisoning: A Stage-wise Benchmarkcited by 0
  5. 2026Generating Multi-view Adversarial Examples for Visual Geometry Grounded Transformercited by 0
  6. 2026A Survey on Adversarial Attacks and Defenses for Diffusion Models Across Multiple Modalitiescited by 0