Even Rockets Cannot Make Pigs Fly Sustainably: Can BGP be Secured with BGPsec?

Qi Li, Yi-Chun Hu, Xinwen Zhang2014 · DOI

TL;DR

This paper argues that BGPsec, a proposed solution to secure the internet's routing protocol (BGP), fundamentally fails to address BGP's inherent security weaknesses. The authors demonstrate that even with BGPsec, critical attacks like prefix hijacking and routing loops remain possible, suggesting a complete redesign of BGP is necessary for robust security.

What Problem Does This Paper Address?

The internet's ability to route traffic relies on the Border Gateway Protocol (BGP). However, BGP has long been plagued by security vulnerabilities, allowing malicious actors to hijack traffic, create blackholes, and disrupt routing. These issues stem from BGP's core design, which lacks built-in mechanisms to verify the authenticity and integrity of routing information. While various security enhancements have been proposed, BGPsec is a recent IETF standard aiming to provide strong security guarantees by authenticating prefix origins and routing paths. This paper questions whether BGPsec can truly solve these deep-seated problems, arguing that its reliance on BGP's existing architecture limits its effectiveness.

Methods

The authors critically analyze BGPsec against four desirable security properties for inter-domain routing: routing availability, path predictability, blackhole-resistant routing, and loop-free routing. They don't propose new attack vectors but rather demonstrate how existing attack classes can still succeed even with BGPsec fully deployed.

Specifically, they:

  • Analyze BGPsec's limitations: They examine how BGPsec's mechanisms, while improving authenticity, do not inherently solve fundamental BGP issues.
  • Illustrate attack scenarios: They detail how wormhole attacks and mole attacks can bypass BGPsec's security measures.
  • Wormhole Attacks: These attacks involve colluding Autonomous Systems (ASes) creating tunnels to make a fake routing path appear shorter and more attractive to victim ASes. The paper shows how BGPsec's path validation can be fooled by these tunnels, as the signatures themselves are valid, even if the path is not legitimate.
  • Mole Attacks: These attacks exploit situations where an AS announces a prefix but doesn't fully utilize it, creating forwarding loops. The authors demonstrate that BGPsec, by focusing on announced paths rather than actual traffic flow and prefix usage, cannot prevent these loops from being exploited to overload network links.
  • Simulate attack impact: They use real-world AS topologies to simulate the effects of wormhole attacks, quantifying the number of hijacked routing paths.

Datasets

The paper uses two measured Internet AS topologies for their simulations:

  • 830-set topology: Derived from the SSFNet project's BGP routing table.
  • rv-set topology: Derived from a CAIDA dataset, focusing on ASes contributing to the Router-Views repository and their neighbors.

These datasets are used to model AS relationships and simulate routing policies.

Key Insights and Ideas

  • BGP's fundamental flaws persist: The core argument is that BGPsec, by building upon BGP's existing architecture, inherits its fundamental security vulnerabilities. It's like putting a fancy lock on a flimsy door – the door itself is still weak.
  • Authenticity vs. Reality: BGPsec can verify that a route announcement is signed by the claimed AS, but it cannot guarantee that the traffic will actually follow that path or that the path is not artificially shortened or looped.
  • Wormholes bypass path validation: Even with BGPsec's path signature verification, colluding ASes can create tunnels that make a fake path appear valid and shorter, leading to traffic hijacking.
  • Mole attacks exploit prefix usage gaps: BGPsec doesn't address the issue of ASes announcing prefixes they don't fully use. This creates opportunities for attackers to induce forwarding loops and disrupt connectivity, as BGPsec cannot verify actual prefix utilization.
  • A complete redesign is likely needed: The paper strongly suggests that incremental fixes like BGPsec are insufficient, and a radical rethinking of how inter-domain routing is secured is required.

Key Results

  • Wormhole attacks remain effective: Simulations on real-world AS topologies showed that wormhole attacks could hijack a significant percentage of routing paths. For instance, in the 830-set graph, scenarios saw hundreds of thousands of routing paths hijacked. The rv-set graph also showed substantial hijacking, with 72% and 44% of nodes experiencing hijacked paths in random and high-attack scenarios, respectively.
  • Mole attacks can create loops: The paper demonstrates that mole attacks can generate permanent forwarding loops, impacting network links and potentially causing widespread disruption. They identified a large number of vulnerable links and target prefixes that could be exploited.
  • BGPsec exacerbates availability issues: The added verification steps in BGPsec can actually increase route convergence time, potentially worsening existing availability problems in BGP.

Limitations and Future Work

  • Limitations:
    • The paper focuses on demonstrating that BGPsec cannot achieve certain security properties, rather than proposing a new secure protocol.
  • The simulations of wormhole attacks are based on inferred AS topologies, which might miss some real-world eBGP links that could affect attack effectiveness.
  • The paper acknowledges that usage-based RPKI certificates could help against mole attacks, but notes the significant operational complexity this would introduce.
  • Future Work:
    • The authors explicitly call for a re-evaluation of BGP's fundamental design principles to achieve strong inter-domain routing security. This implies a need for research into entirely new routing protocols or paradigms.
  • They suggest that network operators might need to implement filters to drop traffic to unused prefixes, but note this is a partial solution and doesn't address the dynamic nature of prefix usage.

Summary updated

Abstract

The Border Gateway Protocol (BGP) suffers from numerous security vulnerabilities, which the BGPsec protocol is supposed to fix.In this paper, we argue that fundamental properties of BGP have inherent security vulnerabilities, and that a complete re-design of BGP is needed to achieve strong security guarantees.
BibTeX
@misc{li2014even,
  author = {Qi Li and Yi-Chun Hu and Xinwen Zhang},
  title = {{Even Rockets Cannot Make Pigs Fly Sustainably: Can BGP be Secured with BGPsec?}},
  year = {2014},
  doi = {10.14722/sent.2014.23001},
}
Figure