Are Your LLM-based Text-to-SQL Models Secure? Exploring SQL Injection via Backdoor Attacks

Summary: Study of backdoor vulnerabilities in LLM Text-to-SQL: ToxicSQL uses stealthy command-like and character-level triggers to hide backdoors while keeping benign accuracy. Demonstrates executable SQL-injection backdoors with 79.41% success using only 0.44% poisoned data and offers defenses. (summarized by gpt-5-mini on Feb 11 2026)

Paper ID: 7548
Venue: SIGMOD
Year: 2026
Pagerank: 5.093636e-05
Overall Rank: 10,344 | 29.04%
DOI: 10.1145/3769762

Incoming Non-self Citations Over Time

No non-self incoming citations found for this paper in this database.

Authors

1. Meiyu Lin (Sichuan University)
2. Haichuan Zhang (Sichuan University)
3. Jiale Lao (Cornell University)
4. Renyuan Li (Sichuan University)
5. Yuanchun Zhou (Chinese Academy of Sciences)
6. Carl Yang (Emory University)
7. Yang Cao (Tokyo Institute of Technology)
8. Mingjie Tang (Sichuan University)

BibTeX Citation

@inproceedings{lin_sigmod26,
        title = {{Are Your LLM-based Text-to-SQL Models Secure? Exploring SQL Injection via Backdoor Attacks}},
        author = {Lin, Meiyu and Zhang, Haichuan and Lao, Jiale and Li, Renyuan and Zhou, Yuanchun and Yang, Carl and Cao, Yang and Tang, Mingjie},
        series = {{SIGMOD} '26},
        booktitle = {Proceedings of the {ACM} {SIGMOD} International Conference on Management of Data},
        publisher = {Association for Computing Machinery},
        doi = {10.1145/3769762},
        url = {https://dl.acm.org/doi/10.1145/3769762},
        year = {2026}
}

Incoming Citations (Sorted by Pagerank)

Showing 0 of 0 citing papers.

Rank	Citing Paper	Year	Venue	Pagerank

Outgoing Citations (Sorted by Pagerank)

Showing 10 of 10 cited papers.

Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.

Rank	Cited Paper	Year	Venue	Pagerank
279	Text-to-SQL Empowered by Large Language Models: A Benchmark Evaluation	2024	VLDB	0.00022468369
756	CodeS: Towards Building Open-source Language Models for Text-to-SQL	2024	SIGMOD	0.0001431656
1,444	CatSQL: Towards Real World Natural Language to SQL Applications	2023	VLDB	0.00010769944
2,039	ScienceBenchmark: A Complex Real-World Benchmark for Evaluating Natural Language to SQL Systems	2024	VLDB	9.2721259e-05
2,710	OpenSearch-SQL: Enhancing Text-to-SQL with Dynamic Few-shot and Consistency Alignment	2025	SIGMOD	8.2206448e-05
2,748	Few-shot Text-to-SQL Translation using Structure and Content Prompt Learning	2023	SIGMOD	8.1707811e-05
2,852	The Dawn of Natural Language to SQL: Are We Fully Ready?	2024	VLDB	8.0455088e-05
3,787	Combining Small Language Models and Large Language Models for Zero-Shot NL2SQL	2024	VLDB	7.1249098e-05
4,251	FinSQL: Model-Agnostic LLMs-based Text-to-SQL Framework for Financial Analysis	2024	SIGMOD	6.8031106e-05
6,936	Reliable Text-to-SQL with Adaptive Abstention	2025	SIGMOD	5.7349543e-05

Semantically Similar Papers

#	Overall Rank	Paper	Year	Venue
1	756	CodeS: Towards Building Open-source Language Models for Text-to-SQL	2024	SIGMOD
2	9,310	LLM-PBE: Assessing Data Privacy in Large Language Models	2024	VLDB
3	6,267	Automated Validating and Fixing of Text-to-SQL Translation with Execution Consistency	2025	SIGMOD
4	2,395	OmniSQL: Synthesizing High-quality Text-to-SQL Data at Scale	2025	VLDB
5	3,787	Combining Small Language Models and Large Language Models for Zero-Shot NL2SQL	2024	VLDB
6	10,732	RTS+: Reliable Text to SQL	2025	SIGMOD
7	6,936	Reliable Text-to-SQL with Adaptive Abstention	2025	SIGMOD
8	9,870	Natural Language to SQL: State of the Art and Open Problems	2025	VLDB
9	10,556	OpenSQL: Data-Efficient Text-to-SQL for Open-Source LLMs via Synthesized Intermediate Supervision	2026	VLDB
10	279	Text-to-SQL Empowered by Large Language Models: A Benchmark Evaluation	2024	VLDB