hi i am researching on Explainable Spatio-Temporal intrusion detection system for Connected and Automated vehicles. can you find me good closed access journals not older than 5 years ago. we need a good base paper which we can counter through our research
1 cited papers · May 1, 2026 · Powered by Researchly AI
The intersection of explainable AI, spatio-temporal modeling, and intrusion detection for Connected and Automated Vehicles (CAVs) is an active and growing resea…
The intersection of explainable AI, spatio-temporal modeling, and intrusion detection for Connected and Automated Vehicles (CAVs) is an active and growing research frontier. Nassif et al. (2021) conducted a systematic review of 290 ML-based anomaly detection papers (2000–2020), identifying 29 distinct ML models and noting that unsupervised anomaly detection has been increasingly adopted by researchers. Choi et al. (2021) further highlight that anomaly detection in multivariate time-series data — directly relevant to CAV sensor streams — poses particular challenges due to the need for simultaneous consideration of temporal dependencies and inter-variable relationships.
- Machine Learning for Anomaly Detection — A broad family of supervised, unsupervised, and semi-supervised techniques applied across 43 identified application domains, with 22 benchmark datasets documented in the literature. Nassif et al. (2021)
- Deep Learning for Time-Series Anomaly Detection — Deep learning models capable of learning large-scale sequence representations in an unsupervised manner, applied to multivariate time-series data with real-world benchmarks. Choi et al. (2021)
- Autonomous Vehicle Control & Architecture — The self-driving system pipeline encompassing perception, planning, and control, with ML techniques applied to vehicle control tasks in connected environments.
- Spatio-Temporal Dependency Modeling — The requirement to simultaneously model temporal sequences and spatial/variable relationships in multivariate data streams, a core challenge for IDS in CAV networks.
CAV Network / V2X Communication Layer | [Sensor & Network Data Streams] (CAN bus, LiDAR, GPS, V2V/V2I) | [Spatio-Temporal Feature Extraction] Temporal: LSTM / Transformer Spatial: GNN / CNN | [Anomaly / Intrusion Detection Model] (Unsupervised / Semi-supervised) | [Explainability Layer] (SHAP, LIME, Attention Maps) | [Alert / Decision Output] → Benign | Attack Type Classified
| Dimension | Nassif et al. (2021) | Choi et al. (2021) |
|---|---|---|
| Scope | Broad ML anomaly detection SLR | Deep learning for time-series anomaly detection |
| Data Type | General (43 application domains) | Multivariate time-series |
| Models Covered | 29 distinct ML models | State-of-the-art deep models |
| Dominant Paradigm | Unsupervised > supervised | Unsupervised representation learning |
| Benchmark Datasets | 22 datasets documented | Multiple benchmark datasets compared |
| Key Gap | Domain-specific deployment not addressed | Models are highly use-case specific; require domain knowledge |
A critical conflicting finding worth noting: Nassif et al. (2021) observe that unsupervised methods dominate adoption , while Choi et al. (2021) caution that such models are highly specific to individual use cases and require domain expertise for appropriate deployment — a tension directly relevant to CAV IDS design.
The retrieved evidence does not include papers specifically addressing explainability (XAI) in CAV intrusion detection systems, which is a significant gap for your research focus. Furthermore, the deep learning time-series review notes that most state-of-the-art models are highly use-case specific and require domain knowledge, limiting their generalizability to CAV threat environments without significant adaptation.
- Unsupervised ML anomaly detection has been the dominant paradigm across 290 reviewed studies, making it a strong baseline to challenge or extend in your CAV IDS work.
- Deep learning models show impressive progress on multivariate time-series anomaly detection but remain highly domain-specific — a key vulnerability your explainable approach can address.
- The CAV system architecture (perception → planning → control) provides the structural context within which your spatio-temporal IDS must operate.
- Benchmark datasets for anomaly detection are well-documented (22+ identified), giving you a foundation for empirical comparison in your research.
- Temporal and spatial dependency modeling simultaneously is the core unsolved challenge in multivariate anomaly detection for connected systems.
- "Explainable intrusion detection system CAN bus autonomous vehicles SHAP 2022–2024" — to find XAI-specific IDS papers for in-vehicle networks
- "Graph Neural Network spatio-temporal anomaly detection V2X connected vehicles IEEE Transactions 2022–2024" — to locate GNN-based spatio-temporal IDS architectures
- "Federated learning intrusion detection connected automated vehicles privacy 2023" — to explore distributed IDS paradigms as a counter-angle to centralized approaches
⚠️ Important Note for Your Research: The retrieved evidence blocks do not contain papers directly on explainable spatio-temporal IDS for CAVs. For your base paper search, I strongly recommend querying IEEE Transactions on Intelligent Transportation Systems, IEEE Transactions on Vehicular Technology, and Computers & Security (all closed-access, high-impact) using the search strings above on IEEE Xplore or Scopus. Papers by Choi et al. (2021) and Nassif et al. (2021) serve as strong methodological baselines you can position your work against.
Research smarter with AI-powered citations
Researchly finds and cites academic papers for any research topic in seconds. Used by students across India.