mlsanityData to 14 Oct 2026

CLIP-guided Diffusion Model for Backdoor Generation in Sensor-based Human Activity Recognition

arXiv 2606.22837 ↗First posted 22 Jun 2026cs.LG cs.AI cs.CRcited by 0

On the map: LLM safety and security → Backdoor and data poisoning attacks

Abstract (arXiv)

Sensors are critical components of modern intelligent devices. The proliferation of the Internet of Things (IoT) and wearable mobile devices has enabled the integration of such sensors to monitor the environment and enable users to take predictive actions. Human activity recognition (HAR) is a popular application in which Inertial Measurement Unit (IMU)-based sensors, such as accelerometers and gyroscopes, are used to provide insights into health, training, and medical diagnosis. However, the accuracy of such a model is hindered by the lack of data. The diffusion model-based technique has proven successful in generating synthetic data for training HAR models. In this paper, we propose a backdoor training technique, IMU-DM-CLIP, that leverages a diffusion model to enable trigger-based attacks on HAR models. Our empirical analysis shows that the attack is successful even with a very small backdoor injection rate of 10% and 10% of the data guided for the diffusion model.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Backdoor and data poisoning attacks, one of 12 directions in LLM safety and security. Builds on 15 of its 32 references that are in the map. First posted on arXiv 22 Jun 2026.

Builds on

15 of 32 references.

Closest ancestors · 2

Same direction, newest first

  1. 2025PCAP-Backdoor: Backdoor Poisoning Generator for Network Traffic in CPS/IoT Environments26 Jan
  2. 2022How to Backdoor Diffusion Models?11 Dec

Other references · 13

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2020Supervised Contrastive Learningcited by 2,829
  3. 2017BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chaincited by 1,106
  4. 2018How To Backdoor Federated Learningcited by 765
  5. 2021DiffusionCLIP: Text-Guided Diffusion Models for Robust Image Manipulationcited by 642
  6. 2019Hidden Trigger Backdoor Attackscited by 371
  7. 2022A Survey on Generative Diffusion Modelcited by 240
  8. 2021Poisoning and Backdooring Contrastive Learningcited by 151
  9. 2021Federated Intrusion Detection for IoT with Heterogeneous Cohort Privacycited by 14
  10. 2025Dynamic User-controllable Privacy-preserving Few-shot Sensing Frameworkcited by 8
  11. 2025Dynamic Black-box Backdoor Attacks on IoT Sensory Datacited by 8
  12. 2025Log Anomaly Detection with Large Language Models via Knowledge-Enriched Fusioncited by 8
1 more
  1. 2026Contrastive Continual Learning for Model Adaptability in Internet of Thingscited by 4

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2023Unsupervised Statistical Feature-Guided Diffusion Model for Sensor-based Human Activity Recognition
  2. 2025Adaptive and Robust Data Poisoning Detection and Sanitization in Wearable IoT Systems using Large Language Models
  3. 2023Cross-Domain HAR: Few Shot Transfer Learning for Human Activity Recognition
  4. 2025Towards Generalizable Human Activity Recognition: A Survey
  5. 2021A Light-weight Deep Human Activity Recognition Algorithm Using Multi-knowledge Distillation
  6. 2022Label Flipping Data Poisoning Attack Against Wearable Human Activity Recognition System
  7. 2024Transformer-Based Approaches for Sensor-Based Human Activity Recognition: Opportunities and Challenges
  8. 2025Refining Diffusion Models for Motion Synthesis with an Acceleration Loss to Generate Realistic IMU Data