FARSIGHT 论文剖析 15 个学术金融 LLM 交易方案的安全与鲁棒性缺陷
Trading Agents or Market Crashers? Dissecting Robustness and Security Failures in Academic Financial LLM Trading Schemes
AI 导读
Mengxiao Wang 与 Nitesh Saxena 发布 FARSIGHT 论文，系统评估学术金融 LLM 交易方案的鲁棒性与安全性，覆盖 15 个方案并对照 7 份开源代码。
推荐理由
论文系统评估学术 LLM 交易智能体的安全缺口，指出代码层面普遍缺失止损与熔断等模型外风控。
正文 · 原文
Abstract:Large language model (LLM) agents already trade real capital, and builders copy a few published designs that share models and news feeds, so one shock reaches many agents at once. We present FARSIGHT (Financial Agent Robustness and Security Investigation and Global Holistic Taxonomy), a systematization that rates trading agents on robustness to market stress and on security against source manipulation, agent compromise, and agents acting as attackers, and checks the ratings against released code. Across 15 academic schemes, no paper describes a halt that skips LLM inference or combines source provenance with cross-source corroboration, and the 7 code releases enforce no stop-loss or position limit outside the LLM. Three LLMs sell whole positions on a fake headline and on real hoaxes, including the fake press release behind the 2000 Emulex crash. In a simulated market driven by these decisions, one fake headline crashes the market in all runs with 400 agents, and 400 agents that sell when prices fall nearly double a 2010-sized crash. The pause that U.S. exchanges apply to large stocks starts in no headline run, whereas automatic halts and hard loss limits outside the model remove both effects. Restated as a domain-independent threat model, FARSIGHT maps the same gaps onto agents in healthcare, law, software, and crypto, where compromised agents spread harm as worms and as cascades. We derive 14 open problems and urge deployers to corroborate sources, keep interrupts that bypass the model, and enforce loss limits outside it.
| Comments: | 57 pages, 19 figures, 21 tables, 221 references. v3: adds a flash-crash experiment (trading decisions of three LLMs on a fake headline and four real market hoaxes, a simulated market with up to 800 agents, the effect of trading halts, hard loss limits, source corroboration, and the exchange Limit Up-Limit Down pause, and robustness checks) and drops the SoK label from the title |
| Subjects: | Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA) |
| ACM classes: | K.6.5; I.2.11; I.2.7; K.4.4; J.1; J.3; K.5 |
| Cite as: | arXiv:2609.19705 [cs.CR] |
| (or arXiv:2609.19705v3 [cs.CR] for this version) | |
| https://doi.org/10.48550/arXiv.2609.19705 arXiv-issued DOI via DataCite |
Submission history
From: Mengxiao Wang [view email]
[v1]
Thu, 17 Sep 2026 04:59:08 UTC (1,044 KB)
[v2]
Thu, 15 Oct 2026 15:20:02 UTC (665 KB)
[v3]
Sun, 25 Oct 2026 20:13:36 UTC (765 KB)
来源：arXiv cs.AI · arxiv.org