Security Economics in the HTTPS Value Chain

Hadi Asghari, Michel van Eeten, Axel Arnbak, N.a.n.m. van EijkSSRN 2013 · DOI

TL;DR

This paper dives deep into the security economics of HTTPS, revealing that despite its critical role in online trust, the system is riddled with systemic vulnerabilities. It argues that the current market for SSL certificates is not a race to the bottom but a highly concentrated space where reputation and bundled services, rather than pure security, drive sales, potentially benefiting dominant players even after breaches.

What Problem Does This Paper Address?

The paper tackles the growing concern that HTTPS, the backbone of secure internet communication, is fundamentally flawed. Despite its widespread adoption and reliance for e-commerce and sensitive transactions, recent high-profile breaches at Certificate Authorities (CAs) have exposed systemic weaknesses. The authors aim to understand why these vulnerabilities persist and how the economic incentives within the HTTPS value chain contribute to this problem, particularly focusing on the role of CAs. They question whether the current market structure encourages genuine security improvements or perpetuates a status quo that benefits established players, even in the face of security failures.

Methods

The research employs a multi-disciplinary approach, combining:

  • Actor-based value chain analysis: Mapping the interactions and dependencies between website operators, CAs, browsers, and end-users to understand the flow of trust and security.
  • Empirical analysis of the SSL certificate market: Utilizing data from the SSL Observatory and market pricing information to analyze market structure, vendor shares, and pricing strategies.
  • Security economics concepts: Applying principles to understand the incentives of CAs and buyers, particularly focusing on information asymmetry and externalities.
  • Desk research and interviews: Gathering insights from existing literature, industry reports, and non-structured interviews with stakeholders.

Datasets

  • SSL Observatory Data: A comprehensive dataset of SSL certificates used to secure HTTPS websites, collected by the Electronic Frontier Foundation (EFF). This dataset, though from December 2010, provided approximately 1.5 million valid SSL certificates.
  • Market Data: A custom dataset mapping CAs to their market names, product offerings, and prices, collected from vendor websites.

Key Insights and Ideas

  • Not a Race to the Bottom: Contrary to common assumptions, the SSL certificate market is not characterized by intense price competition driving down security standards. Instead, it's highly concentrated.
  • Bundled Services and Reputation Drive Sales: The true value proposition in the SSL market isn't just the certificate itself, but bundled services (like enterprise management, liability shields) and the reputation of the CA. This allows for significant price differentiation.
  • "Too Big to Fail" CAs: Market leaders, due to their size and established trust, are perceived as less likely to be removed from browser root stores even after breaches. This "too big to fail" status provides a form of continuity insurance for their customers.
  • Breaches Can Benefit Market Leaders: Ironically, security problems and breaches can increase demand for security signals, benefiting established CAs who are perceived as more reliable, even if their own security practices have been questioned.
  • Weakest Link Problem: The security of the entire HTTPS ecosystem is determined by the weakest CA, a systemic vulnerability that regulatory and technical solutions struggle to fully address.
  • Information Asymmetry and Externalities: Buyers often lack the technical expertise to assess true security, leading them to rely on reputation and perceived security signals. The negative externalities of a CA breach (affecting the entire ecosystem) also complicate incentive alignment.

Key Results

  • The SSL certificate market is highly concentrated, with three vendors (Symantec, GoDaddy, and Comodo) holding over three-quarters of the market share.
  • Significant price differences exist for technically substitutable certificates, with Extended Validation (EV) certificates being the most expensive.
  • High-priced brands, particularly in the EV market, enjoy large market shares, indicating that buyers are willing to pay a premium for perceived security and bundled services.
  • Despite breaches at major CAs like DigiNotar, Comodo, Verisign, and GlobalSign, these companies have largely maintained their market positions, suggesting a resilience to reputation damage.
  • The EU's proposed "eSignatures Regulation" is analyzed, with the authors finding its regulatory solutions potentially insufficient to address systemic vulnerabilities, particularly the weakest-link problem.

Limitations and Future Work

  • Dataset Age: The SSL Observatory data used for market share analysis is from December 2010, creating a time mismatch with the price data collected in February 2013.
  • Scope of Data: The SSL Observatory only captures certificates for publicly visible web servers, excluding digital certificates used for back-end systems or personal email.
  • Price Accuracy: The study acknowledges limitations in price accuracy due to reseller pricing, discounts, and the need for installation on multiple servers.
  • Future Work: The authors suggest further research to test hypotheses about market share changes post-DigiNotar by analyzing more recent SSL Observatory data. They also highlight the need for technical solutions to be adopted at scale and for regulatory proposals to better align incentives and address the weakest-link problem. The paper also calls for extending Mozilla's institutional policy on chain of trust transparency to the entire certificate issuance value chain.

Summary updated

BibTeX
@misc{asghari2013security,
  author = {Hadi Asghari and Michel van Eeten and Axel Arnbak and N.A.N.M. van Eijk},
  title = {{Security Economics in the HTTPS Value Chain}},
  year = {2013},
  doi = {10.2139/ssrn.2277806},
  howpublished = {SSRN},
}

Loading related content...

Figure