Newest papers

Your Token Is Gone: A Token-Consumption Countermeasure for Disrupting LLM-Based Scam Systems

星福 杜, Kailong Wang, Tianlong Yu · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.30666 · cited by 0 works (OpenAlex) · open-access PDF

Spanish

Los sistemas de fraude basados en el contexto de LLM combinan cada vez más el diálogo adaptativo, el contexto persistente, las herramientas externas y la coordinación de múltiples agentes para automatizar el fraude personalizado. Estas capacidades también crean una dependencia de recursos: cada paso de razonamiento, llamada de modelo y ronda de coordinación incurre en un coste de inferencia. Presentamos TokenTrap, una defensa orientada a los recursos que redirecciona los sistemas de fraude sospechosos hacia cálculos benignos de bajo valor, al tiempo que retiene credenciales, pagos y otra información explotable. TokenTrap combina tres componentes: una biblioteca de tareas de token específica para la arquitectura, un router que selecciona entre los flujos de trabajo de agentes únicos, jerárquicos y múltiples, y una estrategia

Machine translation (Meta M2M100); the English original follows.

Abstract

LLM-based scam systems increasingly combine adaptive dialogue, persistent context, external tools, and multi-agent coordination to automate personalized fraud. These capabilities also create a resource dependency: every reasoning step, model call, and coordination round incurs inference cost. We present TokenTrap, a resource-oriented defense that redirects suspected scam systems toward benign, low-value computation while withholding credentials, payments, and other exploitable information. TokenTrap combines three components: an architecture-specific token-task library, a router that selects among single-agent, hierarchical, and flat multi-agent workflows, and a bounded anti-guardrail strategy that varies task presentation while enforcing token, round, time, and safety limits. Across 50-round traces, TokenTrap amplifies token consumption by $2{,}487.0\times$, $1{,}616.0\times$, and $567.8\times$ relative to a no-token-task baseline for the three workflows. Its router achieves 85.67% accuracy on 300 architecture-labeled prompts. These results demonstrate that short, non-sensitive requests can impose substantial context-processing and coordination costs, motivating resource-aware defense as a complement to scam detection.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Security token, Task (project management), Router, Complement (music), Key (lock)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange