Abstract
There have been many incidents of IP prefix hijacking by BGP protocol in the Internet. Attacks may hijack victim’s address space to disrupt network services or perpetrate malicious activities such as spamming and DoS attacks without disclosing identity. The relation between network topology and prefix hijacking influence is presented for all sorts of hijacking events in different Internet layers. The impaction parameter is analyzed for typical prefix hijacking events in different layers. A large Internet emulation environment is constructed and the attack impaction of IP prefix hijacking events are evaluated. The results assert that the hierarchical nature of network influences the prefix hijacking greatly.
Introduction
The Border Gateway Protocol (BGP) is the essential "glue" of the Internet, enabling Autonomous Systems (ASes) to exchange routing information. Despite its critical role, BGP lacks built-in authentication, allowing malicious or misconfigured routers to "hijack" IP prefixes by falsely claiming ownership or offering a "better" path. While detection and prevention are common research topics, the systematic impact of these hijacks—specifically how the positions of the attacker and victim within the global network hierarchy influence the outcome—remains under-explored. This study quantifies this damage to assist network operators in prioritizing defense strategies.
Methodology
The study employs a three-tiered classification of the Internet hierarchy based on commercial relationships:
- Core Layer (C): Top-level providers forming a tightly connected clique.
- Forwarding Layer (F): Middle-tier networks acting as transit points between the core and the edges.
- Marginal/Stub Layer (S): Networks that do not forward data for others, such as small enterprise networks.
Impact Metrics
To evaluate the severity of an attack, two primary metrics are defined:
- $N_c$ (Affected Nodes): The total count of ASes experiencing a routing state change due to the hijack.
- $\mu$ (Affected Path Factor): The percentage of network paths altered by the event.
Emulation Environment
A hybrid emulation environment was constructed using three physical Juniper J2350 routers and four servers. This setup supported 110 ASes, utilizing real-world BGP topology data to ensure the hierarchical layers and commercial relationships mirrored the actual Internet.
Key Results and Analysis
The researchers evaluated nine attack scenarios based on the hierarchical positions of the attacker (A) and the victim (V). The results, summarized in the table below, demonstrate that the network's hierarchical nature is a primary driver of hijack propagation.
Widespread Disruption
The highest number of affected nodes ($N_c = 57$) occurred when the victim was in the Marginal layer and the attacker was in the Forwarding layer. This suggests that "edge" networks are particularly vulnerable to having their traffic diverted globally when targeted by mid-tier providers.
Path Alteration
The most significant path disruption ($\mu = 118$) was observed when the victim was in the Forwarding layer and the attacker was in the Core. Because Forwarding layer networks serve as critical transit points for many other ASes, hijacks targeting them have a disproportionate effect on the stability of global routing paths.
Strategic Insights
- Attacker Height: Within any given level, an attacker positioned higher in the hierarchy (closer to the core) exerts a larger influence over the network.
- Core Resilience: Surprisingly, hijacking a core layer AS is not always the most damaging event in terms of total node count, though it significantly impacts path factors.
- Sub-prefix Hijacks: The study confirms that "sub-prefix" hijacks, where an attacker announces a more specific IP range, are more influential than "same-prefix" hijacks.
Conclusion and Future Work
The hierarchical structure of the Internet plays a decisive role in determining the impact of BGP prefix hijacking. While this study provides a robust framework for understanding these dynamics, it is limited by the scale of the emulation (110 ASes) and the assumption of single-homed ASes. Future research should incorporate multi-homing complexities and larger-scale topologies to further refine these impact models and guide network operator mitigation strategies.
Summary updated