Keys in the Clouds: Auditable Multi-device Access to Cryptographic Credentials

Arseny Kurnikov, Andrew Paverd, Mohammad Mannan, N. Asokan

TL;DR

The authors present a Cloud Key Store (CKS) built on Intel SGX that allows users to securely store and use cryptographic keys across multiple devices without needing specialized hardware like smart cards. By leveraging Trusted Execution Environments (TEEs), the system provides high-performance signing and decryption while introducing novel features like auditable usage logs and time-limited key delegation.

What Problem Does This Paper Address?

Managing personal cryptographic keys (used for PGP, code signing, or file decryption) is a usability and security nightmare.

  • Local storage is bound to a single device and vulnerable to local malware.
  • Hardware tokens (like YubiKeys) are secure but easy to lose and often incompatible with mobile devices.
  • Cloud storage is convenient but traditionally requires you to trust the service provider not to steal your keys or be compromised by hackers.

The authors aim to bridge this gap: providing the "anywhere" access of the cloud with the "hardware-grade" security of a local enclave, all while adding the ability to see exactly when and how your keys are being used.

Methods

The core of the CKS is an Intel SGX enclave. SGX allows a server to set up a "black box" (enclave) where data is encrypted even from the server's own operating system and administrators.

Key technical components include:

  • Remote Attestation: Users verify that the server is running the exact, untampered CKS code before sending any data. They use a streamlined 1-round-trip protocol to minimize setup time.
  • Secure Channel: A Diffie-Hellman key exchange establishes an encrypted tunnel directly between the user's device and the enclave.
  • Rate-Limited Authentication: To protect against password-guessing, the enclave itself enforces an exponential back-off for failed logins. Because this happens inside the TEE, even a malicious server admin cannot bypass the rate limit to brute-force a user's password.
  • Policy-Based Access: The system supports "protected keys" that never leave the enclave. Users can set policies on these keys, such as expiration dates or a maximum number of uses.
  • Integration: The researchers modified GnuPG (Linux) and OpenKeychain (Android) to treat the CKS as a virtual smart card, meaning existing apps can use it without major UI changes.

Datasets

As a systems and security paper, this work does not use training datasets. Instead, it evaluates performance using:

  • Cryptographic Benchmarks: RSA 3072-bit and ECC (secp256r1) operations.
  • Network Simulation: Latency measurements using the CloudPing tool to simulate real-world conditions across various Amazon Web Services (AWS) regions (Ireland, London, Frankfurt).

Key Insights and Ideas

  • The "Aha" of Rate Limiting: Usually, if a server is compromised, an attacker can brute-force passwords offline. By keeping the password check and the rate-limiting logic inside the SGX enclave, the CKS makes even relatively weak passwords resilient against a compromised infrastructure.
  • Auditing as a Prerequisite: The CKS is designed to log every operation before it is performed. This ensures that if an attacker manages to use a key (e.g., via a stolen password), they cannot hide their tracks by deleting the log entry afterward.
  • Delegation without Sharing: Unlike traditional keys where "sharing" means giving someone the actual file, CKS allows you to delegate access. You can let a colleague sign documents with your key for exactly 24 hours, and then revoke that access instantly without ever exposing the underlying private key.

Key Results

  • High Throughput: A single desktop-class CPU can handle approximately 6,000 signature operations per second.
  • Low Latency: Creating a signature via CKS takes about 15ms (locally) to 24ms (via mobile). Even when factoring in 50ms of network lag, it is significantly faster than using a physical smart card, which averaged 1,282ms in their tests.
  • Efficiency: The memory footprint is tiny—about 700 bytes per user record—meaning a standard server could support 100,000 users with just 100MB of RAM.

Limitations and Future Work

  • Side-Channel Attacks: While SGX protects against direct access, it is historically susceptible to "side-channel" attacks (like cache timing). The authors mitigate this by using hardened Intel libraries, but it remains a cat-and-mouse game.
  • Availability & Backups: The paper notes that if a server's hardware clock is tampered with or the enclave state is lost, the keys could become inaccessible. They suggest using consensus protocols (like Raft or Paxos) across multiple enclaves for future production-grade reliability.
  • Legal "Locking": The authors propose a future feature where a user could "lock" their own key for a set period, providing a legal defense that they could not have accessed certain data during that time.

Summary updated

Abstract

Personal cryptographic keys are the foundation of many secure services, but storing these keys securely is a challenge, especially if they are used from multiple devices. Storing keys in a centralized location, like an Internet-accessible server, raises serious security concerns (e.g. server compromise). Hardware-based Trusted Execution Environments (TEEs) are a well-known solution for protecting sensitive data in untrusted environments, and are now becoming available on commodity server platforms. Although the idea of protecting keys using a server-side TEE is straight-forward, in this paper we validate this approach and show that it enables new desirable functionality. We describe the design, implementation, and evaluation of a TEE-based Cloud Key Store (CKS), an online service for securely generating, storing, and using personal cryptographic keys. Using remote attestation, users receive strong assurance about the behaviour of the CKS, and can authenticate themselves using passwords while avoiding typical risks of password-based authentication like password theft or phishing. In addition, this design allows users to i) define policy-based access controls for keys; ii) delegate keys to other CKS users for a specified time and/or a limited number of uses; and iii) audit all key usages via a secure audit log. We have implemented a proof of concept CKS using Intel SGX and integrated this into GnuPG on Linux and OpenKeychain on Android. Our CKS implementation performs approximately 6,000 signature operations per second on a single desktop PC. The latency is in the same order of magnitude as using locally-stored keys, and 20x faster than smart cards.
BibTeX
@inproceedings{kurnikov2018keys,
  author = {Arseny Kurnikov and Andrew Paverd and Mohammad Mannan and N. Asokan},
  title = {{Keys in the Clouds: Auditable Multi-device Access to Cryptographic Credentials}},
  booktitle = {ARES},
  year = {2018},
  url = {https://openreview.net/forum?id=K9uTEz2_qA6},
}

Loading related content...

Figure