Security risks and hierarchical full-domain protection mechanism of large model application in smart education

doi:10.12267/j.issn.2096-5931.2026.08.010

Abstract

Abstract:

Artificial Intelligence (AI) has powerfully driven the transformation and modernization of smart education in higher education. Nevertheless, the rapid proliferation of intelligent technologies has brought prominent data security challenges, which extend far beyond technical defense issues and profoundly affect the stability of teaching operations, the protection of teachers and students’ legitimate rights and interests, and the sustainable innovation of smart education. Focusing on data security practices in university smart education, this paper examines the current construction status and practical dilemmas of smart education, and systematically analyzes the potential security and ethical risks embedded in AI-empowered educational scenarios. Oriented by the principles of security, stability and reliability, this study proposes a systematic safety barrier framework for smart education in institutions of higher education. The research results facilitate the standardized, normative and sustainable development of smart education, and offer feasible implications for improving data security governance in the construction of high-quality university smart education systems.

Key words: smart education, artificial intelligence, data security

CLC Number:

G434

DONG Guangzhi, WU Hanbing, WANG Rui, YOU Jia, WANG Jingyao. Security risks and hierarchical full-domain protection mechanism of large model application in smart education[J]. Information and Communications Technology and Policy, 2026, 52(8): 67-73.

Add to citation manager EndNote|Ris|BibTeX

URL:

http://ictp.caict.ac.cn/EN/10.12267/j.issn.2096-5931.2026.08.010

http://ictp.caict.ac.cn/EN/Y2026/V52/I8/67

Figures/Tables 3

References 10

[1]	刘知远, 孙茂松. 大语言模型幻觉问题研究综述[J]. 计算机学报, 2024, 47 (9): 2017-2042.
[2]	梅宏, 曹锋. 生成式人工智能安全风险与底层技术局限[J]. 软件学报, 2025, 36 (3): 981-1006.
[3]	Kostikova A, Wang Z P, Bajri D, et al. LLLMs:a data-driven survey of evolving research on limitations of large language models[J]. ACM Computing Surveys, 2026, 58 (1): 1-33.
[4]	Alibaba DAMO Academy, NUS. CMM:a benchmark for multimodal hallucination diagnosis[C]//Advances in Neural Information Processing Systems (NeurIPS 2025). Vancouver: NeurIPS, 2025.
[5]	教育部, 国家发展和改革委员会, 科技部, 等. “人工智能+教育”行动计划[EB/OL]. (2026-04-02)[2026-06-01]. http://www.moe.gov.cn/srcsite/A16/s3342/202604/t20260410_1433240.html.
[6]	宋艳飞, 汪慕峰, 周明. 基于人工智能的网络弹性关键技术及发展研究[J]. 工业信息安全, 2025 (3): 27-36.
[7]	薛向阳, 李斌. 数据治理与认知安全[M]. 北京: 科学出版社, 2025.
[8]	俞跃渊. 数据流通交易运营机制研究[J]. 中外企业文化, 2023 (S1): 23-28.
[9]	Selwyn N. What’s theproblem with learning analytics?[J]. Journal of Learning Analytics, 2019, 6 (3): 11-19.
[10]	丁洁. 智慧教育背景下技术与伦理的双向观照与风险消解[J]. 大理大学学报, 2025, 10 (5): 88-92.

防护方案	攻击面覆盖	风险检出准确率	误拦截率	自动化适配能力
静态关键词审核	仅覆盖显性单轮违规词汇,四层底层、长链路、多智能体风险完全漏检	多轮投毒、越权漏洞检出率<28%	正常教学提问易命中关键词,误拦截率>42%	完全自动化
单层大模型安全护栏	无分层知识库约束,缺少多智能体协同防护模块	长链式越狱、知识库投毒漏检率为36.1%	单模型幻觉导致知识点提问误拦截	完全自动化
人工后台巡检	依赖管理员经验,小众协同攻击、隐蔽投毒无法识别	综合检出率为64.78%	人工复核过滤,误拦截率低	无实时自动化拦截能力

防护方案	攻击面覆盖	风险检出准确率	误拦截率	自动化适配能力
静态关键词审核	仅覆盖显性单轮违规词汇,四层底层、长链路、多智能体风险完全漏检	多轮投毒、越权漏洞检出率<28%	正常教学提问易命中关键词,误拦截率>42%	完全自动化
单层大模型安全护栏	无分层知识库约束,缺少多智能体协同防护模块	长链式越狱、知识库投毒漏检率为36.1%	单模型幻觉导致知识点提问误拦截	完全自动化
人工后台巡检	依赖管理员经验,小众协同攻击、隐蔽投毒无法识别	综合检出率为64.78%	人工复核过滤,误拦截率低	无实时自动化拦截能力

防护方案	风险检出准确率	漏检率	误拦截率	单批次全量检测耗时/h
静态关键词审核	56.41%	43.59%	42.17%	0.9
单层大模型安全护栏	77.25%	22.75%	18.63%	2.1
人工后台巡检	64.78%	35.22%	3.46%	11.3
分层全域防护方案	95.82%	4.18%	5.29%	2.8

防护方案	风险检出准确率	漏检率	误拦截率	单批次全量检测耗时/h
静态关键词审核	56.41%	43.59%	42.17%	0.9
单层大模型安全护栏	77.25%	22.75%	18.63%	2.1
人工后台巡检	64.78%	35.22%	3.46%	11.3
分层全域防护方案	95.82%	4.18%	5.29%	2.8

[1]	CAO Haixiao, WANG Yan. Common problems and improvement approaches in data security management capability assessment practice [J]. Information and Communications Technology and Policy, 2026, 52(8): 23-29.
[2]	ZHANG Jiahui, ZHAO Yuhan. Construction of lightweight data security governance system and low-cost implementation path for SMEs [J]. Information and Communications Technology and Policy, 2026, 52(8): 30-36.
[3]	HUANG Jie. Research on criminal law response to attacks against large GenAI models [J]. Information and Communications Technology and Policy, 2026, 52(8): 45-51.
[4]	GUO Xiaotong, XIE Runbin, WU Di, XIA Peng. Research on technology path and architecture of industrial large-scale model for intelligent factory in discrete industry [J]. Information and Communications Technology and Policy, 2026, 52(8): 91-96.
[5]	XU Ming. A review of cross-domain data security governance for low-altitude one-network unified flight platform [J]. Information and Communications Technology and Policy, 2026, 52(6): 70-78.
[6]	BI Ran, LI Fengqiu, GE Jian, HUANG Yongtao. AI computing technologies for intelligent terminals: paths and challenges [J]. Information and Communications Technology and Policy, 2026, 52(3): 14-21.
[7]	CHENG Ying, LIU Zhipeng, ZHANG Meng. Research on development pathways, potential impacts, and governance strategies of terminal intelligent agents [J]. Information and Communications Technology and Policy, 2026, 52(3): 2-7.
[8]	CHEN Xin, ZHANG Hengsheng, FU Tao, LI Guo. Research on the development status and trends of the industrial computing and network convergence industrial artificial intelligence [J]. Information and Communications Technology and Policy, 2026, 52(1): 31-40.
[9]	ZHANG Bin. Driving factors for digital transformation of leadership in the AI era: research based on technology acceptance model [J]. Information and Communications Technology and Policy, 2026, 52(1): 59-67.
[10]	YU Da, DONG Xiaofei, CAO Feng, ZHA Fusheng, SUN Lining. Research on continual learning method for overcoming catastrophic forgetting of artificial neural networks [J]. Information and Communications Technology and Policy, 2026, 52(1): 75-83.
[11]	YAN Jiahao. Current situation and development suggestions for the construction of urban intelligent computing power base [J]. Information and Communications Technology and Policy, 2025, 51(9): 2-7.
[12]	DONG Hao, ZHOU Jingcai. Research on key technologies of artificial intelligence for research and development [J]. Information and Communications Technology and Policy, 2025, 51(8): 42-49.
[13]	LIU Zhipeng, ZHANG Jinming. From platform economy to artificial intelligence: comparison of governance mechanisms and construction of governance systems [J]. Information and Communications Technology and Policy, 2025, 51(8): 55-63.
[14]	QIN Sisi, YAN Dongwei, QI Kexin, CHENG Yang. Research on the evolutionary trends of software forms empowered by artificial intelligence [J]. Information and Communications Technology and Policy, 2025, 51(8): 64-70.
[15]	WU Chongqing, YANG Sizhe, NING Tingyong, HU Yongbo. Exploration of domain-specific large model applications in Shanghai’s digital government [J]. Information and Communications Technology and Policy, 2025, 51(8): 78-83.