Jul 16, 2026 · Artificial Intelligence

Can Overthinking in Large Language Reasoning Models Trigger DoS Attacks? A New Risk Unveiled

Researchers from Zhejiang University and Alibaba Security reveal that large language reasoning models can be forced into excessive, self‑correcting reasoning—'overthinking'—by crafted inputs, dramatically inflating token output and computation cost, enabling a novel black‑box DoS attack demonstrated via a Hierarchical Genetic Algorithm.

DoS attackblack-box attackhierarchical genetic algorithm

0 likes · 10 min read

Can Overthinking in Large Language Reasoning Models Trigger DoS Attacks? A New Risk Unveiled

Machine Learning Algorithms & Natural Language Processing

Jul 11, 2026 · Artificial Intelligence

Overthinking Large Language Models: New DoS Threat to Reasoning Models Unveiled

The paper introduces a black‑box hierarchical genetic algorithm that automatically perturbs the logical structure of reasoning questions to induce excessive chain‑of‑thought in large language models, dramatically inflating output tokens (up to 26.1× on MATH) and creating a DoS‑style resource‑exhaustion attack, with extensive experiments across multiple models demonstrating the vulnerability and its transferability.

DoS attackLLMhierarchical genetic algorithm

0 likes · 10 min read

Overthinking Large Language Models: New DoS Threat to Reasoning Models Unveiled