[ARC] - Risk Parameter Updates for Aave v2 Ethereum - LT and LTV (2022.12.01)

Simple Summary

A proposal to adjust 4 total risk parameters, including Liquidation Threshold (LT) and Loan-To-Value (LTV), for USDC and DAI markets on Aave V2 Ethereum.

Abstract

Following recent market events, the AAVE community has decided to reduce the risk profile across many higher volatile assets by disabling borrow or freezing across v2 markets (AIP-121, AIP-124, and AIP-125)

After mitigating the immediate risk to the protocol with the above AIPs, Chaos Labs is exploring the impact and trade-offs of decreasing the Liquidation Thresholds and LTVs of USDC and DAI. Reducing capital efficiency on v2 assets will mitigate attack vectors while also serving as effective motivation to shift usage to v3, where users can enjoy the benefits of E-mode with enhanced risk techniques and mitigation methodologies.

As Liquidation Threshold reductions may lead to user accounts being eligible for liquidations upon their approval, we want to make the full implications clear to the community at each step. To best minimize this impact, we suggest reaching the desired settings by a series of incremental decreases, following the Risk-Off Framework previously approved by the community, with a reduction of up to 3% in any given AIP. In an attempt to avoid liquidations, Chaos Labs will communicate the planned amendments 7 days before their on-chain execution via all available avenues (Twitter, forums, Nansen Connect, and Blockscan Chat).

Motivation

The priority of this proposal is not to optimize markets (capital efficiency); but rather to focus on the security and risk mitigation of the protocol. It will be followed up with additional proposals to update risk parameters to reflect the current market conditions accurately.

The goal of this ARC is two-fold:

  • Reduce LTs → Reducing liquidation thresholds for USDC and DAI effectively reduces capital efficiency and increases the required capital for waging price manipulation attacks on Aave.
  • Encourage user migrations to v3 → This also begins to pave the way towards migrating to v3, where these assets can be utilized in E-mode.

However, this is a significant change, and we wanted to present data to quantify and visualize the effect of such reductions on protocol users for community discussion. Specifically, we want to surface data around the liquidations this would trigger, as some are sizable and warrant a discussion of how these should be handled and if this is the most appropriate action forward.

USDC

$USDC
Liquidation Threshold 89% → 86%
Loan-To-Value 87% → 80%
Newly unhealthy accounts after reduction 168
Liquidation Value ~1.15M USD
User loss (penalty) ~50K USD

DAI

$DAI
Liquidation Threshold 90% → 87% 90% → 82% 90% → 80%
Loan-To-Value 74% 69% 67%
Newly unhealthy accounts after reduction 11 19 34
Liquidation Value ~1.8K USD 20K USD 850K USD
User loss (penalty) ~70 USD ~800 USD 34K USD

For $DAI, we present several options. We recommend abiding by the Rate of Change framework, which dictates a maximum of a 3% reduced LT with minimal liquidations. However, given the market conditions, we would also like to present several other options that are more assertive and aim to reduce risk parameters in a more significant manner. Reduced capital efficiency on positions will mitigate risk vectors and also serve as an effective first step to encourage migration to v3 post-launch.

Joint Reduction of $USDC and $DAI LTs

Below we present the simulation results of reducing LTs as follows:

  • DAI: 90% → 87%
  • USDC 89% → 86%, respectively.

  • The main item is to focus on accounts at risk that have supplied both USDC and DAI as collateral.
  • 102 accounts in total have provided both $USDC and $DAI as collateral
  • We see that the Liquidatable amount reducing USDC and DAI jointly is the sum of the amount separately.
  • Therefore we can reduce both assets LT jointly without causing more liquidations.

Summary

Based on prior community discourse and market activity over the past few weeks, Aave should continue reviewing avenues to derisk all v2 markets with a specific eye on LTs & LTVs. Forced liquidations of user accounts are a major factor in deciding how and when to reduce these parameters to minimize negative user experience.

The risk parameter options are presented below:

The effect of even small reductions is much higher for USDC than DAI, and the two changes should be viewed separately. We believe that the change to USDC is necessary, but the number of accounts and the amount of value to be liquidated may be too high for immediate action. DAI, on the other hand, can be reduced with minimal user impact and should be strongly considered.

We would appreciate community discussion and feedback on the appropriate path forward for each asset.

6 Likes

I’m generally in favor of this but would personally prefer a more sensitive approach. 1.5M is a lot of liquidations. How would this look like with a 150 bps step instead of 300?

I would assume doing multiple smaller steps would result in less overall liquidations as ppl don’t directly drop below 1hf, but have a chance to rebalance.

2 Likes

Couple questions:

  1. Do you have any estimates of how this will affect VaR in the protocol? This force liquidates 168 users - this is maybe 10x more than have been force liquidated in the entire history of the protocol. While we will be the first people to tell you that lowering LTs is an essential part of risk management, we’d hope that there would be large gains in reduced risk for doing such a drastic reduction.
  2. The first stated goal - reducing the required capital for price manipulation attacks makes a lot of sense. Do you have a model for how much this would reduce the cost? In some models, this would only have a single digit percentage effect on cost, which would probably not deter an attacker very much. One thing to note - the CRV insolvency was not clearly a price manipulation attack and so it’s not obvious that recent market events would necessarily justify these changes.
  3. The second goal is also a good one. However, it would probably make more sense to wait until v3 launches on mainnet to take steps in that direction. So while we’re as excited as anyone for Aave to migrate to v3, unless there is a really large reduction in risk from these changes, I’m not sure if it would make sense to focus on this goal until v3 launch. Or if migration is the main goal - interest rate and fee changes might be a better way to accomplish that.
  4. Any reason we didn’t include ETH here if the goal is to reduce the cost of price manipulation attacks? Couldn’t an attacker just use ETH for the same purpose?

We’re still trying to figure out to work with another risk manager here, but generally support changes that improve the risk/capital efficiency tradeoff, regardless of who posts the ARC. Just would need more info on that tradeoff to support at this point

5 Likes

Happy to see engagement and feedback on the proposal. As we’ve mentioned in the post, the community should decide on the tradeoff between risk appetite and the UX on the platform, in this case, accounts and amounts being liquidated.

Reducing USDC LT by 1.5% to 87.5% would result in 29 accounts eligible for liquidation for a total value of 91K USD.

In the snapshot vote, we plan to include the different options for the community to vote on.

As mentioned, the current goal is not to optimize the tradeoff between capital efficiency and risk but to prioritize economic security and risk mitigation during times of high market volatility. The metrics we’re considering here are:

  • the profitability of a market manipulation attack
  • capital efficiency of market manipulation attacks (i.e., is there an opportunity for an outsized return for an attacker who brings $X capital upfront)

We won’t provide a detailed PnL of the attack, which could be used by malicious actors, but our analysis, based on simulations utilizing the Chaos Asset Protection Tool (amongst others), shows that reducing LTs for selected assets will significantly help reduce the viability of these attack vectors.

We’re eager to collaborate with Gauntlet on these proposals and would love to see your analysis of the impact of these proposals (i.e., adding VaR analysis); we think it’d be extremely impactful for the community.

This is the first proposal of an incoming series aimed at reducing risk parameters to acceptable levels across several assets. We aim to introduce these updates incrementally within DAO-approved bounds.

We believe the current risk parameters are high at the moment and should be reduced to meet the goal of keeping Aave v2 open for business. We want to be as risk-off as possible (without freezing all markets) as v3 is on the horizon and a better alternative for users and the protocol as a whole.

3 Likes

Due to the difference in the magnitude of LT reductions between USDC and DAI, we’ve decided to hold two separate Snapshot votes, which will be live in 24 hours from December 7th, 6 PM EST to December 10th, 6 PM EST. Following these votes, we will create a joint AIP incorporating the Snapshot results.

Links to Snapshot vote:

The tables below provide up-to-date data on the implications of LT reductions of 150 and 300 bps for the two assets and will be the basis of this proposal.

DAI

$DAI Current Option 1 Option 2
Liquidation Threshold 90% 88.5% 87%
Loan-To-Value 77% 75% 75%
Newly unhealthy accounts after reduction 10 11
Liquidation Value ~1.2K USD ~1.56K USD
User loss (penalty) ~50 USD ~60 USD

USDC

$USDC Current Option 1 Option 2
Liquidation Threshold 89% 87.5% 86%
Loan-To-Value 87% 80% 80%
Newly unhealthy accounts after reduction 37 94
Liquidation Value ~82.69K USD ~1.16M USD
User loss (penalty) ~3.7K USD ~52K USD

Given the data above, Chaos Labs’ recommendation is for an initial reduction of:

  • DAI - 300 bps (Option 2)
  • USDC - 150 bps (Option 1)
1 Like

Following the snapshot votes (USDC, DAI), we plan to submit an AIP in the coming days with the following suggested amendments:

$USDC
LT 89% → 87.5%
LTV 87% → 80%
$DAI
LT 90% → 87%
LTV 77% → 75%

To minimize liquidations, we will update this thread daily (until the proposal is executed) with a list of accounts at risk of being liquidated due to the change. Please note that given changes in market prices, the number of accounts on this list may change daily.
In addition, we will be utilizing Nansen Connect and Blockscan Chat to message accounts directly.

1 Like

As of today, 51 total accounts amounting to ~$88K are at risk of liquidation following the proposed changes.
The full list of accounts and their projected health score, given the updated LTs can be found in this sheet.

1 Like

We have published this AIP for community members to vote on:

Thank you in advance for your participation in the vote.

Accounts in risk of liquidation:
As of today, 53 total accounts amounting to ~$88K are at risk of liquidation following the proposed changes.
The full list of accounts and their projected health score, given the updated LTs can be found in this sheet.

As of today, 61 total accounts amounting to ~$99K are at risk of liquidation following the proposed changes.
The full list of accounts and their projected health score, given the updated LTs can be found in this sheet .

As of today, 58 total accounts amounting to ~$90K are at risk of liquidation following the proposed changes.
The full list of accounts and their projected health score, given the updated LTs can be found in this sheet .

As of today, 55 total accounts amounting to ~$88.7K are at risk of liquidation following the proposed changes.
The full list of accounts and their projected health score, given the updated LTs can be found in this sheet .

As of today, 60 total accounts amounting to ~$89.1K are at risk of liquidation following the proposed changes.
The full list of accounts and their projected health score, given the updated LTs can be found in this sheet .