Can the proof of elasped time (PoET) consensus algorithm be exploited by multiple nodes?

Question

i recently learned about the Hyperledger Blockchain Project and the specific frameworks. In case of the Sawtooth framework, consensus is reached not by PoW but by PoET (Proof of elapsed Time).

When getting familiar with the proof-of-work (PoW) you usually hear that it's safe regarding participant exploits, because of the computation power which would be needed.

In regards of PoET where the leader for making the consensus is elected by the node with the shortest wait time, i am wondering if it would not be possible to exploit the system, when spaming the network with multiple nodes, so it's more likely to become the leader.

BR

Answer 1

Yes, here's an article stating that "SgxSpectre Attack Can Extract Data from Intel SGX Enclaves".

According to researchers, SgxSpectre works because of specific code patterns in software libraries that allow developers to add SGX support to their apps. Vulnerable SGX development kits include the Intel SGX SDK, Rust-SGX, and Graphene-SGX.

Academics say an attacker can leverage the repetitive code execution patterns that these SDKs introduce in SGX enclaves and watch for small variations of cache size. This is a classic "side-channel attack," and is quite effective.

However, Intel stated that an update for the Intel SGX SDK that adds SgxSpectre mitigations will be released on March 16, 2018.

Also, Here's a research paper that does some theoretical analysis on compromising a blockchain network using SgX and finally states that "Yes, it might be possible".

The title of this paper is "On Security Analysis of Proof-of-Elapsed-Time (PoET)". I'm adding relevant snips of the paper to ensure that in case dead links occur, at least the summary of what they say is available in this post.

The brief of this paper states that

"To fulfil this gap, we develop a theoretical framework for evaluating a PoET based blockchain system, and show that the current design is vulnerable in the sense that an adversary can jeopardize the blockchain system by only compromising Θ(log log n/ log n) fraction of the participating nodes, which is very small when n is relatively large.

Based on our theoretical analysis, we also propose methods to mitigate these vulnerabilities"

They go on to summarise that there are ways to mitigate these problems.

• Changing the probability distribution of F. If the probability distribution F does not rely on n, then adversaries have to compromise a a+µ fraction of the nodes in order to compromise the system.

  • Allowing blocks generated by honest mining nodes to be rejected. We assume that the statistical test will not reject a block that is generated by an honest node, whereas the adversaries can simulate the fastest honest node in the system. It is possible to get beyond the threshold of Θ(log log n log n) if we allow the statistical test to reject blocks generated by honest users.

However, it is important to note that this paper does not mention a concrete method to compromise the safety of the PoET algorithm running in an SGX enclave and merely states that

"SGX and other trusted computing technologies are not 100% reliable. Especially, they may be vulnerable to sophisticated adversaries with necessary resources and skillsets."

But given the SGXSpectre attack, it is not an academic question but a practical one ....

UPDATE: Here's an excellent video titled "34C3 - Microarchitectural Attacks on Trusted Execution Environments" where Keegan Ryan explains in great detail about how TEE attacks are performed. This is absolute gold for a person interested in defending against security breaches.

TLDR: Yes, it is indeed possible to compromise the PoET algorithm (SGXSpectre attack) as also by compromising a small fraction of the participant nodes. There are also some ways to mitigate if not resolve this issue (SGX SDK Patch, other approaches).

Answer 2

No. Sawtooth is a CPU-agnostic blockchain platform. It includes an optional TEE/SGX feature which enhances BFT protections for PoET. PoET is designed following a defense-in-depth approach. There are three or so mechanisms that work in different aspects of the protocol independently from the TEE. This includes three tests performed by PoET:

• c-test: A node must wait c blocks after admission before its blocks will be accepted - this is to prevent trying to game identities and some obscure corner sc enarios.
• K-test: The node can publish at most K blocks before its peers require it to recertify itself.
• z-test: And perhaps most importantly a node may not publish at frequency greater than z

Finally, should a node run a compromised consensus protocol, the main characteristic at risk would be fairness. It would not be able to impact correctness network-wide. That is, it cannot publish invalid transactions. If it does the other nodes will just reject those transactions and the associated block(s) and they will not commit network-wide.