Surviving the Unseen: Predictive Defense for Novel Multi-Turn Multimodal Attacks

Abstract

The expansion of Multimodal Large Language Models (MLLMs) and their integration into autonomous agentic workflows has introduced a non-stationary attack surface. Empirical observations indicate that adversaries employ progressive, cross-modal perturbations that evade turn-specific guardrails by distributing malicious intent across longitudinal conversational trajectories. Static defense mechanisms, constrained by the Markov property, evaluate inputs in isolation and fail to detect cumulative str

Categories

@misc{you2026surviving,
  title = {{Surviving the Unseen: Predictive Defense for Novel Multi-Turn Multimodal Attacks}},
  author = {Doohee You},
  year = {2026},
  month = may,
  eprint = {2605.18988},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2605.18988}
}