TL;DR
This paper introduces arpsec, a novel system that secures the Address Resolution Protocol (ARP) without requiring any changes to the protocol itself. By combining a logic-based prover with Trusted Platform Module (TPM) attestations, arpsec provides a robust defense against ARP spoofing and cache poisoning attacks with minimal performance overhead.
What Problem Does This Paper Address?
The Address Resolution Protocol (ARP) is a fundamental component of IPv4 networking, responsible for mapping IP addresses to MAC addresses. However, ARP is notoriously vulnerable to attacks like spoofing and cache poisoning. These attacks can enable more sophisticated threats such as Man-in-the-Middle (MitM) and Denial-of-Service (DoS) attacks, compromising network security and user privacy. While previous attempts to secure ARP have often involved modifying the protocol itself or requiring complex infrastructure, this paper addresses the need for a secure ARP solution that is compatible with existing systems, efficient, and trustworthy.
Methods
arpsec employs a two-tiered approach for validating ARP replies:
- Logic Layer: This layer uses a logic prover to reason about the validity of ARP bindings. It maintains a history of trusted address bindings and uses a set of formal logic rules to determine if a new binding is authoritative. This layer is designed to be lightweight and provide a first line of defense.
- TPM Layer: If the logic layer cannot definitively validate a binding,
arpsecleverages the Trusted Platform Module (TPM) present in most modern computers. It implements an Attestation (AT) protocol where the local machine challenges the remote machine. The remote machine's TPM provides a signed quote of its system state (PCR values) and an Attestation Identity Key (AIK) for authentication. This allowsarpsecto verify the integrity and identity of the remote host.
Crucially, arpsec operates in user space and does not require any modifications to the core ARP protocol or the Linux kernel, making it easier to deploy.
Datasets
Not discussed in the paper.
Key Insights and Ideas
- Protocol-Agnostic Security: The most significant insight is that ARP security can be achieved without altering the underlying ARP protocol. This is a critical factor for widespread adoption, as modifying existing network infrastructure is often a major hurdle.
- Logic + Hardware Trust: Combining a formal logic prover with hardware-based TPM attestations creates a powerful and flexible security mechanism. The logic layer handles common cases efficiently, while the TPM provides a strong, tamper-resistant root of trust for more complex or suspicious scenarios.
- Amortized TPM Costs: While TPM attestations can be slow,
arpsec's design allows these costs to be amortized. The logic layer handles most validations, meaning TPM attestations are only invoked when necessary, and their cost can be spread over many ARP operations. - Trustworthiness Beyond Binding:
arpsec's TPM integration emphasizes that simply having a correct MAC/IP binding isn't enough; the trustworthiness of the system providing that binding is paramount.
Key Results
- Minimal Overhead:
arpsecintroduces a system overhead ranging from 7% to 15.4% compared to standard Linux ARP, which is competitive with or better than existing secure ARP proposals like S-ARP and TARP. - Efficiency: The logic layer is designed to be efficient, and the integration of the Prolog engine with the C-based daemon achieved a 50x performance improvement.
- TPM Bottleneck: The primary performance bottleneck is the TPM hardware itself, particularly the quote operation, which can take up to 336 ms. However, this is mitigated by the logic layer and amortized over time.
- Comparison to S-ARP/TARP:
arpsecdemonstrates lower overhead than S-ARP and comparable or better overhead than TARP, especially in worst-case scenarios (e.g., when the ARP cache is flushed).
Limitations and Future Work
- TPM Hardware Dependency:
arpsecrelies on the presence of TPM hardware. While TPMs are common, legacy systems may lack them, requiring software TPM solutions which themselves need secure execution environments. - DoS/DDoS Defense:
arpsecis not designed to defend against sophisticated DoS/DDoS attacks, though it can mitigate some simple DoS scenarios through blacklisting. - Future Work: The authors suggest extending
arpsecto support Gratuitous ARP and IPv6 Neighbor Discovery Protocol (NDP), as the core concepts are applicable. They also mention integrating with integrity measurement architectures like IMA or PRIMA for further TCB reduction and exploring dynamic root of trust environments.
Summary updated