Paper uses game theory and LLMs to bait phone scammers
Original title: Game of Phones: Harnessing Game Theory and LLMs in 'Bot Wars' to Counteract Phone Scams
Across 300 simulated dialogues, adaptive play was 28.9–79.7% closer to Nash equilibrium than static baselines.
A cs.CR paper models phone scammer and scam-baiter interaction as a two-player game and lets LLMs pick counter-strategies from in-context experience without fine-tuning. Nash equilibrium is the benchmark. Across 300 simulated dialogues and three LLM architectures, adaptive selection was 28.9–79.7% closer to equilibrium than non-adaptive baselines. Code and an annotated set of 300 dialogues with turn-level strategy labels and utility scores are on GitHub. The evaluation is simulated, not a deployed defense.
Unconfirmed: Results are from simulated dialogues, not live engagement with real phone scammers.
- Existing countermeasures impose no resource cost on perpetrators, the authors argue.
- Strategies and utility functions are described as empirically grounded.
- Results cover three unnamed LLM architectures.
- The dataset includes turn-level strategy labels and utility scores.
From the original article
1 min read at arxiv.orgPhone scams impose substantial and growing costs on individuals and financial systems worldwide, yet existing countermeasures impose no resource cost on perpetrators. Strategic scam-baiting, in which automated agents engage scammers to deplete their operational capacity, offers a proactive alternative but requires dynamic strategy adaptation that static systems cannot provide.
By Nardine Basta · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.