A Neurosymbolic Framework for Interpretable Cognitive Attack Detection in Augmented Reality

TL;DR

CADAR is a neurosymbolic framework for detecting cognitive attacks in Augmented Reality — attacks that alter AR content to manipulate what users perceive. It fuses neural vision-language perception with particle-filter-based statistical reasoning, inheriting the adaptability of pre-trained VLMs and the interpretability of Bayesian filtering.

Key contributions:

• A symbolic perception-graph representation fused from multimodal vision-language inputs, incorporating prior knowledge, salience weighting, and temporal correlations.
• Particle-filter based statistical reasoning (sequential Monte Carlo) over the perception graph for interpretable, rigorous attack detection.
• Accuracy improvements of up to 10.7% over strong baselines on an extended AR cognitive-attack dataset.

BibTeX

@misc{chen2025neurosymbolicframeworkinterpretablecognitive,
      title={A Neurosymbolic Framework for Interpretable Cognitive Attack Detection in Augmented Reality}, 
      author={Rongqian Chen and Allison Andreyev and Yanming Xiu and Mahdi Imani and Bin Li and Maria Gorlatova and Gang Tan and Tian Lan},
      year={2025},
      eprint={2508.09185},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2508.09185}, 
}