README.md

Weakly-supervised Anomaly Detection: A Survey

This repo is constructed for collecting and categorizing papers about weakly supervised anomaly detection models according to our survey paper——Weakly Supervised Anomaly Detection: A Survey

Summary and categorization of weakly supervised anomaly detection (WSAD) algorithms

We first summarize and further categorize existing WSAD algorithms into three categories, including: (i) incomplete supervision; (ii) inexact supervision; (iii) inaccurate supervision

1.Summary of WSAD Algorithms

Method	Reference	Venue	Backbone	Modalities	Key Idea	Official Code
			Incomplete Supervision
OE	ref	KDD'14	-	Tabular	Anomaly feature representation learning	×
XGBOD	ref	IJCNN'18	-	Tabular	Anomaly feature representation learning	√
DeepSAD	ref	ICLR'20	MLP	Tabular	Anomaly feature representation learning	√
ESAD	ref	BMVC'21	MLP	Tabular	Anomaly feature representation learning	×
DSSAD	ref	ICASSP'21	CNN	Image/Video	Anomaly feature representation learning	×
REPEN	ref	KDD'18	MLP	Tabular	Anomaly feature representation learning	×
AA-BiGAN	ref	IJCAI'22	GAN	Tabular	Anomaly feature representation learning	√
Dual-MGAN	ref	TKDD'22	GAN	Tabular	Anomaly feature representation learning	√
SemiGNN	ref	ICDM'19	MLP+Attention	Graph	Anomaly feature representation learning	×
ACT	ref	AAAI'23	GNN	Graph	Anomaly feature representation learning	√
DevNet	ref	KDD'19	MLP	Tabular	Anomaly score learning	√
FEAWAD	ref	TNNLS'21	AE	Tabular	Anomaly score learning	√
LEDGM	ref	TNNLS'21	AE	Tabular	Anomaly score learning	×
PReNet	ref	KDD'23	MLP	Tabular	Anomaly score learning	√
Overlap	ref	KDD'23	-	Tabular	Anomaly score learning	√
TargAD	ref	ICDE'24	AE	Tabular	Anomaly score learning	√
SAD	ref	IJCAI'23	AE	Graph	Anomaly score learning	√
SNARE	ref	KDD'09	-	Graph	Graph learning and label propagation	×
AESOP	ref	KDD'14	-	Graph	Graph learning and label propagation	×
SemiGAD	ref	IJCNN'21	GNN	Graph	Graph learning and label propagation	×
Meta-GDN	ref	WWW'21	GNN	Graph	Graph learning and label propagation	√
SSAD	ref	JAIR'13	-	Tabular	Active learning	×
AAD	ref	ICDM'16	-	Tabular	Active learning	√
SLA-VAE	ref	WWW'22	VAE	Time series	Active learning	×
SOEL	ref	ICML'23	CNN	Tabular/Image	Active learning	√
Meta-AAD	ref	ICDM'20	MLP	Tabular	Reinforcement learning	√
DPLAN	ref	KDD'21	MLP	Tabular	Reinforcement learning	×
CutAddPaste	ref	KDD'24	CNN	Time-series	Data Augmentation	√
NNG-Mix	ref	TNNLS'24	-	Tabular	Data Augmentation	√
RoSAS	ref	IP&M'23	MLP	Tabular	Data Augmentation	√
ADGym	ref	NIPS'23	-	Tabular	AutoML	√
ConsisGAD	ref	ICLR'24	GNN	Graph	Data Augmentation	√
GenGA	ref	KDD'24	GCN	Graph	Data Augmentation	×
			Inexact Supervision
Sultani et al.	ref	CVPR'18	MLP	Video	Multiple Instance Learning	√
TCN-IBL	ref	ICIP'19	CNN	Video	Multiple Instance Learning	×
AR-Net	ref	ICME'20	MLP	Video	Multiple Instance Learning	√
RTFM	ref	ICCV'21	CNN+Attention	Video	Multiple Instance Learning	√
Zhu et al.	ref	BMVC'19	AE+Attention	Video	Multiple Instance Learning	×
Purwanto et al.	ref	ICCV'21	TRN+Attention	Video	Multiple Instance Learning	×
MPRF	ref	IJCAI'21	MLP+Attention	Video	Multiple Instance Learning	×
MCR	ref	ICME'22	MLP+Attention	Video	Multiple Instance Learning	×
XEL	ref	SPL'21	MLP	Video	Cross-epoch Learning	√
MIST	ref	CVPR'21	MLP+Attention	Video	Multiple Instance Learning	√
MSLNet	ref	AAAI'22	Transformer	Video	Multiple Instance Learning	√
SRF	ref	SPL'20	MLP	Video	Self Reasoning	×
VadCLIP	ref	AAAI'24	Attention	Video	Pre-trained Foundation Model	√
MGFN	ref	AAAI'23	CNN+Attention	Video	Multiple Instance Learning	√
TPWNG	ref	CVPR'24	Attention	Video	Pre-trained Foundation Model	-
CoMo	ref	CVPR'23	GCN+CNN	Video	Multiple Instance Learning	-
UMIL	ref	CVPR'23	MLP	Video	Multiple Instance Learning	√
UR-DMU	ref	AAAI'23	Attention	Video	Multiple Instance Learning	√
PUMA	ref	KDD'23	AE	Time-series	Multiple Instance Learning	√
CU-Net	ref	CVPR'23	MLP	Video	Self-training	√
CLAWS Net+	ref	TNNLS'23	MLP	Video	Clustering	√
WETAS	ref	ICCV'21	MLP	Time-series/Video	Dynamic Time Warping	×
Iwata et al.	ref	ML Journal'20	AE	Tabular	AUC maximization	×
Isudra	ref	TIST'21	-	Time-series	Bayesian optimization	√
			Inaccurate Supervision
LAC	ref	CIKM'21	MLP/GBDT	Tabular	Ensemble learning	×
ADMoE	ref	AAAI'23	Agnostic	Tabular	Ensemble learning	√
BGPAD	ref	ICNP'21	LSTM+Attention	Time series	Denoising network	√
SemiADC	ref	IS Journal'21	GAN	Graph	Denoising network	×
NRGL	ref	IJCAI'24	MLP	Graph	Denoising network	√
Zhong et al.	ref	CVPR'19	GCN	Video	Problem Transformation	√

2.Categorization of WSAD algorithms

2.1 AD with Incomplete Supervision

• Anomaly Feature Representation Learning

  • OE
    📄Learning outlier ensembles:The best of both worlds–supervised and unsupervised
  • XGBOD
    📄Xgbod: improving supervised outlier detection with unsupervised representation learning
    👉Code Link
  • DeepSAD
    📄Deep semi-supervised anomaly detection
    👉Code Link
  • ESAD
    📄Esad: End-to-end deep semi-supervised anomaly detection
  • REPEN
    📄Learning representations of ultrahigh-dimensional data for random distance-based outlier detection
  • DSSAD
    📄Learning discriminative features for semi-supervised anomaly detection
  • AA-BiGAN
    📄Anomaly detection by leveraging incomplete anomalous knowledge with anomaly-aware bidirectional gans
    👉Code Link
  • Dual-MGAN
    📄Dual-mgan: An efficient approach for semi-supervised outlier detection with few identified anomalies
    👉Code Link
  • SemiGNN
    📄A Semi-Supervised Graph Attentive Network for Financial Fraud Detection\
  • ACT
    📄Cross-Domain Graph Anomaly Detection via Anomaly-aware Contrastive Alignment
    👉Code Link

• Anomaly Score Learning

  • DevNet
    📄Deep anomaly detection with deviation networks
    👉Code Link
  • PReNet
    📄Deep weakly-supervised anomaly detection 👉Code Link
  • FEAWAD
    📄Feature encoding with autoencoders for weakly supervised anomaly detection
    👉Code Link
  • LEDGM
    📄Semisupervised Training of Deep Generative Models for High-Dimensional Anomaly Detection\
  • Overlap
    📄Anomaly Detection with Score Distribution Discrimination
    👉Code Link
  • TargAD
    📄A Robust Prioritized Anomaly Detection when Not All Anomalies are of Primary Interest
    👉Code Link
  • SAD
    📄SAD: Semi-Supervised Anomaly Detection on Dynamic Graphs
    👉Code Link

• Graph Learning

  • SNARE
    📄Snare: a link analytic system for graph labeling and risk detection

  • AESOP
    📄Guilt by association: large scale malware detection by mining file-relation graphs

  • SemiGNN
    📄A semi-supervised graph attentive network for financial fraud detection

  • SemiGAD
    📄Semi-supervised anomaly detection on attributed graphs

  • Meta-GDN
    📄Few-shot network anomaly detection via cross-network meta-learning
    👉Code Link

  • SemiADC
    📄Semi-supervised anomaly detection in dynamic communication networks

  • GraphUCB
    📄Interactive anomaly detection on attributed networks
    👉Code Link

• Active learning and reinforcement learning

  • SSAD
    📄Toward supervised anomaly detection
  • AAD
    📄Incorporating expert feedback into active anomaly discover
    👉Code Link
  • Meta-AAD
    📄Meta-aad: Active anomaly detection with deep reinforcement learning
    👉Code Link
  • DPLAN
    📄Toward deep supervised anomaly detection: Reinforcement learning from partially labeled anomaly data
  • SLA-VAE
    📄A semi-supervised vae based active anomaly detection framework in multivariate time series for online systems
  • SOEL
    📄Deep Anomaly Detection under Labeling Budget Constraints
    👉Code Link

• Data Augmentation

  • CutAddPaste
    📄CutAddPaste: Time Series Anomaly Detection by Exploiting Abnormal Knowledge 👉Code Link
  • NNG-Mix
    📄NNG-Mix: Improving Semi-supervised Anomaly Detection with Pseudo-anomaly Generation
    👉Code Link
  • ROSAS
    📄RoSAS: Deep Semi-supervised Anomaly Detection with Contamination-resilient Continuous Supervision
    👉Code Link
  • ADGym
    📄ADGym: Design Choices for Deep Anomaly Detection
    👉Code Link
  • ConsisGAD
    📄Consistency Training with Learnable Data Augmentation for Graph Anomaly Detection with Limited Supervision
    👉Code Link
  • GenGA
    📄Graph Anomaly Detection with Few Labels: A Data-Centric Approach\

2.2 AD with Inexact Supervision

• MIL-based
  • Sultani et al.
    📄Real-world anomaly detection in surveillance videos
    👉Code Link
  • AR-Net
    📄Weakly supervised video anomaly detection via center-guided discriminative learning
    👉Code Link
  • TCN-IBL
    📄Temporal convolutional network with complementary inner bag loss for weakly supervised anomaly detection
  • RTFM
    📄Weakly-supervised video anomaly detection with robust temporal feature magnitude learning
    👉Code Link
  • Zhu et al.
    📄Motion-aware feature for improved video anomaly detection
  • Purwanto et al.
    📄Dance with self-attention: A new look of conditional random fields on anomaly detection in videos
  • MPRF
    📄Weakly-supervised spatio-temporal anomaly detection in surveillance video
  • MCR
    📄Multi-scale continuity-aware refinement network for weakly supervised video anomaly detection
  • XEL
    📄Cross-epoch learning for weakly supervised anomaly detection in surveillance videos
    👉Code Link
  • MIST
    📄MIST: Multiple instance self-training framework for video anomaly detection
    👉Code Link
  • MSLNet
    📄Self-training multi-sequence learning with transformer for weakly supervised video anomaly detection
    👉Code Link
  • VadCLIP
    📄VadCLIP: Adapting Vision-Language Models for Weakly Supervised Video Anomaly Detection
    👉Code Link
  • MGFN
    📄MGFN: Magnitude-Contrastive Glance-and-Focus Network for Weakly-Supervised Video Anomaly Detection
    👉Code Link
  • TPWNG
    📄Text Prompt with Normality Guidance for Weakly Supervised Video Anomaly Detection\
  • CoMo
    📄Look Around for Anomalies: Weakly-supervised Anomaly Detection via Context-Motion Relational Learning\
  • UMIL
    📄Unbiased Multiple Instance Learning for Weakly Supervised Video Anomaly Detection
    👉Code Link
  • UR-DMU
    📄Dual Memory Units with Uncertainty Regulation for Weakly Supervised Video Anomaly Detection
    👉Code Link
  • PUMA
    📄Learning from Positive and Unlabeled Multi-Instance Bags in Anomaly Detection
    👉Code Link
  • CU-Net
    📄Exploiting Completeness and Uncertainty of Pseudo Labels for Weakly Supervised Video Anomaly Detection
    👉Code Link
• Non MIL-based
  • SRF
    📄A self-reasoning framework for anomaly detection using video-level labels
  • WETAS
    📄Weakly supervised temporal anomaly seg- mentation with dynamic time warping
  • CLAWS Net+
    📄Clustering Aided Weakly Supervised Training to Detect Anomalous Events in Surveillance Videos
    👉Code Link
  • Inexact AUC
    📄Anomaly detection with inexact labels
  • Isudra
    📄Indirectly supervised anomaly detection of clinically meaningful health events from smart home data
    👉Code Link

2.3 AD with Inaccurate Supervision

• Ensemble Learning

  • LAC
    📄Fraud Detection under Multi-Sourced Extremely Noisy Annotations
  • ADMoE
    📄ADMoE: Anomaly detection with mixture-of-experts from noisy labels
    👉Code Link

• Denosing Network

  • BGPAD
    📄Isp self-operated bgp anomaly detection based on weakly supervised learning
    👉Code Link
  • SemiADC
    📄Semi-supervised anomaly detection in dynamic communication networks
  • NRGL
    📄Robust Heterophilic Graph Learning against Label Noise for Anomaly Detection
    👉Code Link

• Problem Transformation

  • Zhong et al.
    📄Graph Convolutional Label Noise Cleaner: Train a Plug-And-Play Action Classifier for Anomaly Detection
    👉Code Link

Experiment

The experimental results can be seen in our previous work ADBench