Interactive convolution and Transformer based anomaly detection algorithm

doi:10.3969/j.issn.1671-1122.2026.07.006

Abstract

Abstract:

To address the limitations of the Transformer model’s attention mechanism in handling the inherent noise and redundant information in time series data, this paper proposed an anomaly detection algorithm based on interactive convolution and Transformer, termed ICT-AD. The algorithm employed an adaptive frequency filter (AFF) to remove high-frequency noise and redundant information, and adopted Anomaly Transformer as its backbone network. Additionally, an adaptive interactive convolution (AIC) block was introduced to further enhance the model’s ability to capture and interpret complex temporal patterns. Extensive experimental results on four datasets demonstrate that ICT-AD outperforms existing methods.

Key words: time series, anomaly detection, Transformer, interactive convolution

CLC Number:

TP309

Zheng Tianming, Liu Shangdong, Li Haitian, Li Hua. Interactive convolution and Transformer based anomaly detection algorithm[J]. Netinfo Security, 2026, 26(7): 1077-1086.

Figures/Tables 4

References 26

[1]	Xu Jiehui, Wu Haixu, Wang Jianmin, et al. Anomaly transformer: time series anomaly detection with association discrepancy[EB/OL]. (2022-04-25)[2025-08-12]. https://arxiv.org/abs/2110.02642.
[2]	Zhang Tianping, Zhang Yizhuo, Cao Wei, et al. Less is more: fast multivariate time series forecasting with light sampling-oriented mlp structures[EB/OL]. (2022-07-04)[2025-08-12]. https://arxiv.org/abs/2207.01186.
[3]	Sellam A Z, Benaissa I, Taleb-ahmed A, et al. Mamba adaptive anomaly transformer with association discrepancy for time series[J]. Engineering Applications of Artificial Intelligence, 2025, 160(5): 11-16.
[4]	Liu Qingxiang, Liu Chenghao, Sun Sheng, et al. GDformer: going beyond subsequence isolation for multivariate time series anomaly detection[EB/OL]. (2025-01-30)[2025-08-12]. https://arxiv.org/abs/2501.18196.
[5]	Yue Wenzhen, Ying Xianghua, Guo Ruohao, et al. Sub-adjacent transformer: improving time series anomaly detection with reconstruction error from sub-adjacent neighborhoods[EB/OL]. (2024-08-03)[2025-08-12]. https://arxiv.org/abs/2404.18948.
[6]	Song J, Kim K, Oh J, et al. MEMTO: memory-guided transformer for multivariate time series anomaly detection[J]. Advances in Neural Information Processing Systems, 2023, 36: 57947-57963.
[7]	Eldele E, Ragab M, Chen Z, et al. TSLANet: rethinking transformers for time series representation learning[EB/OL]. (2024-03-02)[2025-08-12]. https://proceedings.mlr.press/v235/eldele24a.html.
[8]	Zeng Ailing, Chen Muxi, Zhang Lei, et al. Are transformers effective for time series forecasting?[C]// The 37th AAAI Conference on Artificial Intelligence. Menlo Park: AAAI, 2023, 37(9): 11121-11128.
[9]	Cui Manjun, Zhang Zhichao. Generalized convolution and product theorems associated with the free metaplectic transformation and their applications[J]. Digital Signal Processing, 2024, 145(2): 104350-104351. doi: 10.1016/j.dsp.2023.104350 URL
[10]	Omuya E O, Okeyo G O, Kimwele M W. Feature selection for classification using principal component analysis and information gain[J]. Expert Systems with Applications, 2021, 174(18): 27-30
[11]	Shin Y, Lee S, Tariq S, et al. ITAD: integrative tensor-based anomaly detection system for reducing false positives of satellite systems[C]// The 29th ACM International Conference on Information & Knowledge Management. New York: ACM, 2020: 2733-2740.
[12]	Shen Lifeng, Li Zhuocong, Kwok J. Timeseries anomaly detection using temporal hierarchical one-class network[J]. Advances in Neural Information Processing Systems, 2020, 33(22): 13016-13026.
[13]	Ruff L, Vandermeulen R, Goernitz N, et al. Deep one-class classification[C]// The International Conference on Machine Learning. Stockholm: PMLR, 2018: 4393-4402.
[14]	Zong Bo, Song Qi, Min M R, et al. Deep autoencoding gaussian mixture model for unsupervised anomaly detection[C]// The International Conference on Learning Representations. San Diego: ICLR, 2018: 1-19.
[15]	Yairi T, Takeishi N, Oda T, et al. A data-driven health monitoring method for satellite housekeeping data based on probabilistic clustering and dimensionality reduction[J]. Transactions on Aerospace and Electronic Systems, 2017, 53(3): 1384-1401. doi: 10.1109/TAES.2017.2671247 URL
[16]	Breunig M M, Kriegel H P, Ng R T, et al. LOF: identifying density-based local outliers[C]// ACM SIGMOD International Conference on Management of Data. New York: ACM, 2000: 93-104.
[17]	Tariq S, Lee S, Shin Y, et al. Detecting anomalies in space using multivariate convolutional LSTM with mixtures of probabilistic PCA[C]// The International Conference on Multimedia. New York: ACM, 2019: 2123-2133.
[18]	Hundman K, Constantinou V, Laporte C, et al. Detecting spacecraft anomalies using LSTMS and nonparametric dynamic thresholding[C]// The 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. New York: ACM, 2018: 387-395.
[19]	Park D, Hoshi Y, Kemp C C. A multimodal anomaly detector for robot-assisted feeding using an LSTM-based variational autoencoder[J]. Robotics and Automation Letters, 2018, 3(3): 1544-1551.
[20]	Su Ya, Zhao Youjian, Niu Chenhao, et al. Robust anomaly detection for multivariate time series through stochastic recurrent neural network[C]// The International Conference on Multimedia. New York: ACM, 2019: 2828-2837.
[21]	Tuli S, Casale G, Jennings N R. TRANAD: deep transformer networks for anomaly detection in multivariate time series data[EB/OL]. (2022-05-14)[2025-08-12]. https://arxiv.org/abs/2201.07284.
[22]	Wu Qiang, Yao Gechang, Feng Zhixi, et al. PERI-Midformer: periodic pyramid transformer for time series analysis[J]. Advances in Neural Information Processing Systems, 2024, 37(11): 13035-13073.
[23]	Zhang Zhenhua, Li Leyao, Mu Lingxia, et al. Anomaly detection using bidirectional prediction and dynamic thresholding for a spacecraft[C]// The 37th Chinese Control and Decision Conference. New York: IEEE, 2025: 2327-2332.
[24]	Abdulaal A, Liu Zhuanghua, Lancewicki T. Practical approach to asynchronous multivariate time series anomaly detection and localization[C]// ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York: ACM, 2021: 2485-2494.
[25]	Entekhabi D, Njoku E G, O’neill P E, et al. The soil moisture active passive (smap) mission[J]. Proceedings of the IEEE, 2010, 98(5): 704-716. doi: 10.1109/JPROC.2010.2043918 URL
[26]	Goh J, Adepu S, Junejo K N, et al. A dataset to support research in the design of secure water treatment systems[C]// The International Conference on Critical Information Infrastructures Security. Heidelberg: Springer, 2017: 88-99.

模型	MSL			PSM
模型	P	R	F1	P	R	F1
DAGMM	89.60%	63.93%	74.62%	93.49%	70.03%	80.08%
MMPCACD	81.42%	61.31%	69.95%	76.26%	78.35%	77.29%
LOF	47.72%	85.25%	61.18%	57.89%	90.49%	70.61%
ITAD	69.44%	84.09%	76.07%	72.80%	64.02%	68.13%
THOC	88.45%	90.97%	89.69%	88.14%	90.99%	89.54%
Deep-SVDD	91.92%	76.63%	83.58%	95.41%	86.49%	90.73%
CL-MPPCA	73.71%	88.54%	80.44%	56.02%	99.93%	71.80%
LSTM	85.45%	82.50%	83.95%	76.93%	89.64%	82.80%
LSTM-VAE	85.49%	79.94%	82.62%	73.62%	89.92%	80.96%
OmniAnomaly	89.02%	86.37%	87.67%	88.39%	74.46%	80.83%
TSLANet	77.46%	90.12%	83.32%	98.36%	98.55%	97.73%
Peri-midFormer	89.66%	75.31%	81.83%	98.40%	96.01%	97.19%
TranAD	90.38%	99.99%	94.94%	—	—	—
AnomalyTrans	92.09%	95.15%	93.59%	96.91%	98.90%	97.89%
ICT-AD	92.14%	98.07%	95.01%	97.39%	99.14%	98.26%
模型	SMAP			SWaT
模型	P	R	F1	P	R	F1
DAGMM	86.45%	56.73%	68.51%	89.92%	57.84%	70.40%
MMPCACD	88.61%	75.84%	81.73%	82.52%	68.29%	74.73%
LOF	58.93%	56.33%	57.60%	72.15%	65.43%	68.62%
ITAD	82.42%	66.89%	73.85%	63.13%	52.08%	57.08%
THOC	92.06%	89.34%	90.68%	83.94%	86.36%	85.13%
Deep-SVDD	89.93%	56.02%	69.04%	80.42%	84.45%	82.39%
CL-MPPCA	86.13%	63.16%	72.88%	76.78%	81.50%	79.07%
LSTM	89.41%	78.13%	83.39%	86.15%	83.27%	84.69%
LSTM-VAE	92.20%	67.75%	78.10%	76.00%	89.50%	82.20%
OmniAnomaly	92.49%	81.99%	86.92%	81.42%	84.30%	82.83%
TSLANet	92.45%	64.47%	75.96%	91.50%	94.14%	92.80%
Peri-midFormer	90.40%	56.10%	68.62%	91.91%	94.95%	93.40%
TranAD	80.43%	99.99%	89.15%	97.60%	69.97%	84.91%
AnomalyTrans	94.13%	99.40%	96.69%	91.55%	96.73%	94.07%
ICT-AD	94.63%	98.99%	96.76%	94.58%	96.75%	95.65%

模型	MSL	PSM	SMAP	SWaT
Baseline	93.59%	97.89%	96.69%	94.07%
Baseline+AFF	94.31%	98.02%	96.49%	94.60%
Baseline+AIC	94.99%	97.94%	96.48%	94.90%
Baseline+AFF+AIC	95.01%	98.26%	96.76%	95.65%