Space Information Networks (SINs) represent a paradigm shift in global connectivity, integrating satellites (LEO, MEO, GEO), Unmanned Aerial Vehicles (UAVs), and ground stations to support high-dynamic, real-time broadband transmission. However, as these networks become increasingly integrated into terrestrial systems, their vulnerability to cyber-attacks grows. This paper provides a comprehensive survey of the security landscape for SINs, focusing specifically on secure routing and anomaly detection through the lens of both traditional methods and emerging Artificial Intelligence (AI) technologies.
The Unique Security Landscape of SINs
Unlike terrestrial networks, SINs operate in a high-latency, resource-constrained environment with asymmetrical link capacities and highly dynamic topologies. These characteristics render traditional security models insufficient. The authors identify a critical gap: while satellite communication is established, the "integration" of these networks exposes them to sophisticated threats such as eavesdropping, spoofing, and DDoS attacks.
Taxonomy of Security Threats
The paper classifies SIN threats into three primary categories: Natural, Environmental, and Mission threats. Mission threats are particularly concerning, involving malicious activities like the Turla malware or "Zombie invasion" hacks aimed at controlling or destroying network assets.
Secure Routing and Anomaly Detection
Routing in SINs is analyzed across three domains: User-to-Satellite (UDL), Inter-Satellite (ISL), and Boundary routing. The survey compares traditional single-layer strategies against multi-layer intelligent algorithms.
For anomaly detection, the authors propose a novel framework combining a Cybersecurity Knowledge Graph with a Graph Neural Network (GNN)-based Encoder-Decoder.
- Encoder: Embeds graph vertices to capture structural and attribute features.
- Decoder: Reconstructs network topology and attributes.
- Anomaly Scoring: Nodes with a high reconstruction error (D-value) are flagged as potential threats.
This approach treats SINs as Delay Tolerant Networks (DTNs), shifting the focus from constant connectivity to reliable delivery over time. By using GNNs, the system can extract features directly from the moving graph structure of the satellite constellation.
Key Results and Performance Analysis
A major finding of the survey is the significant resilience advantage of multi-layered routing (LEO/MEO/GEO) over single-layer LEO systems. During "oscillatory phases"—periods of high utilization where routes must leap to higher entities—multi-layer systems maintain much higher stability.
As shown in the performance data, Multi-Layered Satellite Routing (MLSR) keeps packet loss below 6%, whereas single-layer LEO routing experiences spikes up to 34% during peak utilization. Furthermore, AI-based anomaly detection methods like ITAD and OCSVM achieved accuracies between 0.88 and 0.99 on satellite-specific datasets such as House Keeping TeleMetry (HKTM).
Future Directions and Challenges
Despite the promise of AI, several bottlenecks remain:
- Onboard Constraints: Limited computer capacity on satellites hinders real-time deployment of heavy deep learning models.
- Lack of Standards: Most current protocols rely on empirical analysis rather than formal security models.
- Emerging Tech: Future research should explore Blockchain for decentralized identity management and Mobile Edge Computing (MEC) to offload security computations to gateways.
Ultimately, the integration of AI, digital twins, and blockchain into SINs will be a long-term technical challenge essential for maintaining cybersecurity in the next generation of global communications.
Summary updated