Noninvasive Methods for Host Certification

Patrick Traynor, Michael Chien, Scott Weaver, Boniface Hicks, Patrick D. McDanielACM Trans. Inf. Syst. Secur 2008 · DOI

TL;DR

This paper introduces a novel, non-invasive method for verifying that a client machine has up-to-date anti-virus software before granting network access. By leveraging zero-knowledge proof principles, clients can prove their security posture without revealing sensitive system details, addressing privacy concerns inherent in traditional scanning methods.

What Problem Does This Paper Address?

In today's interconnected world, granting network access to unknown or potentially compromised devices is a significant security challenge. Traditional methods like user authentication alone are insufficient, as even trusted users can unknowingly host malware. Network administrators often resort to invasive system scans to check for security vulnerabilities and ensure proper functioning of security software like anti-virus. However, these scans raise serious privacy concerns, especially in public access environments like Wi-Fi hotspots, where users may be unwilling to grant arbitrary access to their personal data. This paper tackles the problem of host certification by developing a method that allows clients to prove they have adequate security without compromising their privacy.

Methods

The core of the proposed solution is a "client puzzle" protocol inspired by zero-knowledge proofs. Here's how it works:

  1. Challenge: When a client requests network access, the server sends a "test vector." This vector is a collection of small files, some containing actual (but harmlessly delivered) malware and others containing benign "placebos."
  2. Response: The client's anti-virus software, automatically triggered by the POP3 protocol (or similar), scans these files. The client then reports back to the server, indicating which files it identified as malicious.
  3. Verification: The server compares the client's response to its pre-calculated answer. If the client correctly identifies the malcode, it demonstrates that its anti-virus software is functioning and up-to-date. The client is then granted access.

This approach is non-invasive because the server never directly accesses the client's system files or internal state. The client proves the presence and proper functioning of its security infrastructure through its ability to correctly classify the provided files. The protocol is designed to be agnostic to specific anti-virus software suites, relying on the common feature of automatic scanning of incoming network traffic.

Datasets

Not discussed in the paper. The paper focuses on a protocol and its implementation using existing anti-virus software and simulated malcode.

Key Insights and Ideas

  • Privacy-Preserving Security Verification: The most significant insight is that it's possible to verify security compliance without invasive scans, directly addressing a major user privacy concern.
  • Zero-Knowledge Proof Analogy: The paper effectively uses the concept of zero-knowledge proofs to explain how a client can demonstrate knowledge (of its anti-virus status) without revealing the underlying secrets (its system's internal state).
  • Leveraging Existing Infrastructure: The protocol cleverly exploits the default behavior of anti-virus software (scanning incoming traffic via protocols like POP3) rather than requiring new, specialized client-side software.
  • "Client Puzzle" as a Certification Tool: The concept of a "client puzzle" is a powerful metaphor and mechanism for proving security posture in a verifiable, yet non-intrusive, manner.
  • Assurance Levels: The paper quantifies the level of assurance provided by the protocol using "nines" (e.g., three-nines, nine-nines), directly linking the number of test files to the probability of a correct guess by an uninformed client.

Key Results

  • Efficiency: The microbenchmarks show that the dominant factor in the protocol's execution time is the anti-virus scanning itself, which accounts for over 98.5% of the total time. The protocol's overhead (network transmission, MIME parsing, hashing) is relatively low.
  • Assurance: The protocol can achieve high levels of assurance. For example, using 30 files in a test vector provides nine-nines (0.999999999) probability of assurance. The paper provides an equation to determine the number of files (P) needed for a desired assurance level.
  • Scalability: The implementation is efficient, with the client program occupying minimal disk space and requiring no persistent on-disk or memory footprint.
  • Performance: Macrobenchmarks indicate that the time to achieve nine-nines assurance increases linearly with the number of rounds, but the primary cost remains the anti-virus scanning startup for each round.

Limitations and Future Work

  • Threat Model: The protocol is designed to protect against "unwitting" users who may have outdated or disabled anti-virus software, not against determined, malicious adversaries.
  • Malcode Database Management: A significant challenge for administrators is the maintenance of large, up-to-date databases of malcode for the test vectors. The paper acknowledges that such repositories could potentially be misused.
  • Continued Compliance: The protocol verifies the client's security at the point of entry. There's no guarantee the client will maintain this security posture (e.g., by disabling anti-virus later). The authors suggest a "ticket-based extension" to address this, where clients would periodically re-certify.
  • Generalization: While demonstrated for anti-virus, the authors suggest the approach can be generalized to certify other security services (e.g., open port checks, cryptographic algorithm implementations) by creating appropriate "puzzles."
  • Malcode Selection: The paper discusses strategies for selecting malcode based on time, virulence, and specific services, but this remains a complex area.
  • "Inoculated" Malcode: The authors mention the potential for using non-malicious test viruses (like EICAR) as a safer alternative, drawing parallels to biological vaccines, but note that research in this area is still nascent.

Summary updated

BibTeX
@article{traynor2008noninvasive,
  author = {Patrick Traynor and Michael Chien and Scott Weaver and Boniface Hicks and Patrick D. McDaniel},
  title = {{Noninvasive Methods for Host Certification}},
  journal = {ACM Trans. Inf. Syst. Secur.},
  year = {2008},
  doi = {10.1145/1341731.1341737},
}

Loading related content...

Figure