SMTrap: Cost-Effective DoS Attacks Against Large Reasoning Models via SMT Conflict Guidance
Researchers have developed a new method for launching denial-of-service (DoS) attacks against large reasoning models. The approach, called SMTrap, uses a conflict count derived from an Satisfiability Modulo Theories (SMT) solver to guide the synthesis of inference-heavy queries without relying on model feedback or training an attack model. This allows for more effective and efficient DoS attacks compared to existing methods.
Researchers have developed a new method for launching denial-of-service (DoS) attacks against large reasoning models. The approach, called SMTrap, uses a conflict count derived from an Satisfiability Modulo Theories (SMT) solver to guide the synthesis of inference-heavy queries without relying on model feedback or training an attack model. This allows for more effective and efficient DoS attacks compared to existing methods.
---
Why it matters: SMTrap matters because it enables more powerful and cost-effective DoS attacks against large reasoning models, which could have significant implications for AI system security and reliability.
Source: https://arxiv.org/abs/2608.18921
This article was originally published at: https://arxiv.org/abs/2608.18921