Provenance-gated tool invocation for defending LLM agents against indirect prompt injection

Indirect prompt injection allows the tool-calling agent to execute calls on untrusted data, and the tool whitelist does not review the source of parameters. This paper introduces provenance gate (PGTI), which sequentially performs tool whitelist and parameter source taint checks before the tool execution, allowing the model to still see all tools. In four AgentDojo tasks, 100 attack pairs, and 60 benign tasks, Qwen2.5 14B ran 960 times. The attack success rate decreased from 17.0% with no defense to 3.0%, and it still held after switching models and attack templates; the utility under attack decreased from 40.0% to 28.0%. The benign utility relative to static tool filtering (35.0% and 26.7%) and the increment over the whitelist alone showed no significant difference in this sample. Pair comparison simultaneously reported both safety and utility.

Authors

Institutions

Publication Details

Journal
Advances in Engineering Innovation
Published
2026-11-16
DOI
https://doi.org/10.54254/2977-3903/2026.39089
Primary Topic
Adversarial Robustness in Machine Learning
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Provenance-gated tool invocation for defending LLM agents against indirect prompt injection

Fan Si, Zhongmiao Sun, Lulu Jing, Chao He et al.
1 links
Advances in Engineering Innovation
Adversarial Robustness in Machine Learning
article

Provenance-gated tool invocation for defending LLM agents against indirect prompt injection

Fan Si, Zhongmiao Sun, Lulu Jing, Chao He, Nanmei Chen
article en

Abstract

Indirect prompt injection allows the tool-calling agent to execute calls on untrusted data, and the tool whitelist does not review the source of parameters. This paper introduces provenance gate (PGTI), which sequentially performs tool whitelist and parameter source taint checks before the tool execution, allowing the model to still see all tools. In four AgentDojo tasks, 100 attack pairs, and 60 benign tasks, Qwen2.5 14B ran 960 times. The attack success rate decreased from 17.0% with no defense to 3.0%, and it still held after switching models and attack templates; the utility under attack decreased from 40.0% to 28.0%. The benign utility relative to static tool filtering (35.0% and 26.7%) and the increment over the whitelist alone showed no significant difference in this sample. Pair comparison simultaneously reported both safety and utility.

Advances in Engineering InnovationVol. 17(12)
Beijing Union University (CN), Yonsei University (KR), Peking University (CN), Kyung Hee University (KR), Shanghai Maritime University (CN)
Adversarial Robustness in Machine Learning
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon