Position Matters: Feature Inversion Attacks in ViT Split Inference with Token Reduction and Shuffling
cs.CR, cs.AI
Submitted: 2026-09-01
Updated: 2026-09-17
Comments: Accepted at the 19th ACM Workshop on Artificial Intelligence and Security (AISec'26)
License: http://creativecommons.org/licenses/by/4.0/
The gist: Vision Transformers (ViTs) are increasingly used in split-inference systems, where edge devices transmit intermediate token representations to a remote cloud.
Terminology
Abstract
Vision Transformers (ViTs) are increasingly used in split-inference systems, where edge devices transmit intermediate token representations to a remote cloud. In this setting, token reduction lowers computation and communication costs, while token shuffling disrupts the spatial organization of the transmitted tokens, potentially limiting information leakage. However, their privacy benefits remain unclear against feature inversion attacks, which attempt to reconstruct the input from the transmitted embeddings. In this work, we show that, despite disrupting the spatial structure required by conventional reconstruction attacks, transmitted token embeddings retain substantial positional information. Based on this observation, we introduce the Spatially Aligned Reconstruction Attack (SARA), a unified pipeline that predicts token positions, restores their spatial layout, reconstructs missing embeddings using a feature-space masked autoencoder, and recovers the input image. Our results demonstrate that token shuffling provides only apparent privacy, as SARA largely reconstructs the original token organization. Token reduction offers stronger protection, but significant leakage persists when the retained tokens preserve sufficient semantic and positional information. Finally, we introduce a lightweight edge-side defense that removes positional embeddings and progressively adapts the edge-side transformer blocks through knowledge distillation. It substantially reduces attack performance against SARA, while preserving downstream task accuracy and requiring no changes to the cloud-side model.
Sources
- From Pixels to Privacy: Temporally Consistent Video Anonymization via Token Pruning for Privacy Preserving Action Recognition
- SoK: Let the Privacy Games Begin! A Unified Treatment of Data Inference Privacy in Machine Learning
- DINOv3
- Split learning for health: Distributed deep learning without sharing raw patient data
- Permutation Equivariance of Transformers and Its Applications
- Defending Model Inversion and Membership Inference Attacks via Prediction Purification
- Adaptive Token Merging for Efficient Transformer Semantic Communication at the Edge
- On the (In-)Security of the Shuffling Defense in the Transformer Secure Inference
- A Survey of Visual Transformers
Related papers
- SoK: AI-Augmented Binary Reversing
- Relaxed Sender Anonymity for CBDC Interbank Settlement: A Zero-Knowledge Approach on Permissioned EVM
- Calibration-Family Overfit: Why Trusted Sabotage Monitors Don't Transfer Across Lineages
- Efficient Fuzzy PSI under One-Sided Assumptions
- Sealing the Audit-Runtime Gap for LLM Skills
- Token Composition: A Graph Based on EVM Logs