Redesign Periodic Followee Confirmation

TL:DR
The Periodic Followee Confirmation should be redesigned to fulfill it’s intended function (removing the followee configuration of a neuron, should it not confirm within 6 months), rather than existing for the purpose of removing voting power from the quorum & reducing the security threshold of the network (without even removing the followee configuration of a neuron).

The Periodic Followee Confirmation was a topic first brought to light April 2022, which was passed in proposal #55651.

The concept was pursued by the community, yet pushed off by DFINTY until August of 2024, when the following redesign of periodic followee confirmation was proposed;

This redesign differed from the initial proposed implementation, and raised the following concern;

Receiving the following response;

While there was further debate that can be referenced, this summarizes the direction of the conversation.

In January of 2025, the feature was released;

With the first week resulting in the following voting power removed from the consensus;

A week later, the topic was brought up once more;

Leading to the following math being provided;

While there are still a few days before we see the effects of this first tranche conclude, we currently have a total voting power of 389,499,419, meaning 107,671,080 voting power (53,835,540 ICP / 326,781,727.8$, under the assumption all are 8 year neurons, and without calculating age bonus) has been removed from the consensus.

As a result of this voting power being removed from consensus, DFINITYs percentage of actuated voting power has increased to 25.53% from the previous 20.03% (+27.5%), while DFINITYs actuated voting power with followees has increased from 34.1% to 42% on motion proposals. In addition, the weight of votes cast by active neurons has been increased by 27.6%.

The downwards trend of this tranche should continue for another week, and is likely to repeat every 6 months, although it’s reasonable to assume future tranches will be much less dramatic than the initial.

This is a long way of saying, this system was not designed as initially proposed, and pushed for by the community, and in it’s current form stands to artificially inflate the voting power of participants by eliminating the voting power of others, which is a cannibalistic system.

Regardless as to whether a neuron is voting or not (as the penalty for not voting is not receiving voting rewards), all neurons should contribute to the consensus & security threshold for the fact that it is staked ICP (someones money that is locked) securing the network, which is increasing the quorum to pass proposal & cost of attack by a bad actor, internally or externally.

The Periodic Followee Confirmation should be redesigned to fulfill it’s intended function (removing the followee configuration of a neuron, should it not confirm within 6 months), rather than existing for the purpose of removing voting power from the quorum & reducing the security threshold of the network (without even removing the followee configuration of a neuron).

7 Likes

For those like me that don’t understand the above, this is a translation

# Current Implementation vs. Suggested Redesign: Security Impact

## Current Implementation
### How it works:
- If a neuron (staked ICP) doesn’t vote, set a followee, or confirm its followee within 5 months, its voting power decreases. After 6 months of inactivity, it drops to zero.
- This removed voting power is excluded from the quorum (total voting power for proposals) and security threshold (stake securing the network against attacks, like a 51% attack).
- Example: In the first week, 45M voting power was removed, growing to 107M by March 27, 2025, cutting total voting power from ~497M to 389M.

### Effect on Security:
- **Reduced Total Stake:** Excluding inactive neurons shrinks the effective stake. A 51% attack needs over half the active voting power, so a drop from 497M to 389M means an attacker needs only 194.5M ICP instead of 248.5M.
- **Inflated Active Power:** Active neurons (e.g., DFINITY’s) gain proportional influence (e.g., DFINITY’s share rose from 20% to 25.5%, up to 42% with followees), concentrating power without extra stake.
- **Security Risk:** Lower total voting power reduces the ICP needed for a majority, making attacks easier and cheaper for internal or external bad actors.

## User’s Suggested Redesign
### How it works:
- PFC only removes a neuron’s followee configuration after 6 months of inactivity (per the 2022 proposal), not its voting power.
- All staked ICP, voting or not, contributes to the quorum and security threshold. Inactive neurons lose followee delegation but keep their stake’s weight.
- Penalty for not voting is missing rewards, not exclusion from consensus.

### Effect on Security:
- **Full Stake Counts:** All staked ICP remains in the pool. With 497M staked, the quorum and security threshold stay at 497M, regardless of activity.
- **Stable Attack Cost:** A 51% attack requires over half the total stake (248.5M ICP), not just active power, keeping the attack bar high and costly.
- **Balanced Power:** Active neurons don’t gain inflated influence, as inactive stake dilutes their share. DFINITY’s power wouldn’t jump from 20% to 25.5% without adding stake.

## Key Difference in Security Terms
- **Current System:** Shrinks effective stake, lowers the 51% attack threshold (less ICP needed), and boosts active players’ control, potentially centralizing power and reducing security.
- **Suggested Redesign:** Keeps full stake active, maintains a higher attack threshold (more ICP needed), and avoids power shifts, enhancing security.

## Why the Difference Matters
The user argues staked ICP is locked value securing the network and should always count, not just when voting. The current setup cuts this “defensive” stake, weakening security, while the redesign preserves it, aligning with the original intent: manage followees, not undermine the system.
2 Likes

I completely agree with the @Accumulating.icp .

It’s clear that change is necessary, under the current model, as more stakers become inactive, DFINITY’s voting power continues to grow.

I think we all agree that this shouldn’t be the intended outcome.

2 Likes

I actually think it’s much more likely for neuron holders to reactivate their rewards than for future neuron holders to also become inactive, and that you’re worried about nothing here.

The people who haven’t logged on in 6 months are losing out, and I think that’s fine. Good for the price, and who cares because they’re not paying the slightest bit of attention.

So you’re worried about people who have already logged in, confirmed their neurons, are aware of the 6 month rule just forgetting they did it?

2 Likes

I think he is just saying that the price to do a 51% attack is decreasing ( and over react about it because he think its possible for someone to gather such power )

I think this is a possibility, although I believe it’s just as likely of a possibility that we will see participation rates align with governance, which only sees 65% participation, which would translate to a total voting power of ~325,000,000. In time, as the network grows, this will correct itself, although I don’t think we should be leaving consensus vulnerable in hopes of users returning.

We agree on the fact that if you haven’t logged in for 6 months, you should not be receiving voting rewards, which will contribute towards counteracting inflation, but this system takes it a step further by removing their voting power completely rather than just removing their followees

Removing 175m Voting Power / 87,500,000 ICP / 525,000,000$ from the consensus for the fact that it will benefit certain people.

It makes it more feasible to attack, and with 60m ICP in exchanges, 100m maturity ready to be spawned, and neuron markets (which DFINITY previously preached against for this reason), I don’t think it’s a risk we need to take against DFINITYs ~50m staked ICP.

Nobody had a problem with executing proposals to begin with.

3 Likes

If we think ICP is the future, why do we ignore real world business strategies - just an (extreme) example.

The fix is straightforward: periodic confirmations should remove followees from inactive stakers, not their voting power.

The goal was to reduce inflation, so it’s unclear why DFINITY chose to take it a step further.

The current model increases the risk of centralization by amplifying DFINITY’s relative voting influence.

What was the reasoning behind that? Maybe I’m missing something.

5 Likes

This breakdown made the post much more clear for me. I finally see what you’ve been trying to explain to me for about a year now🍻

1 Like

DFINITY is concerned that if a proposal cannot be passed quickly with an absolute majority, it will be unable to fix critical vulnerabilities in time.

1 Like

If DFINITY must implement a fast absolute majority for certain types of proposals, it would be better to separate them and authorize them through a motion. In practice, there is no real difference. I also don’t think that a group of neurons funded by DFINITY would make the network more decentralized.

1 Like

Following DOES get removed after a long period of inaction. This has always been part of the plan, and is in fact implemented; there simply has not been enough time yet for it to kick in, but it will within a week or so.

1 Like

Great! So that means that solving this is as simple as disabling the function that removes neurons voting power.

2 Likes

I hear what you are saying, but what does the limit look like? Death/abandonment/loss of access causes more and more ‘dead’ voting power to accumulate over time. Even in non-emergencies, this doesn’t seem ideal.

One protection against a 51% attack would simply be a higher price point for ICP, but that happens organically (and would happen the moment someone begins to accumulate a significant amount). It seems your strategy is to artificially deflate the value of ICP as a voting asset by treating inactive neurons the same as active ones.

In business meetings we normally require a quorum for decisions. Let’s say we have 5 members, so our quorum is 3. If 2 members are constantly missing from the meetings, we might eventually add 2 new members, but those members who didn’t show up are going to be removed so that our quorum remains manageable. We aren’t going to raise the quorum to 4 in order to accommodate the 2 members who never vote, because you can imagine how things would eventually turn out for that business.

I think the core problem is that people can follow other neurons, but they typically choose to follow DFINITY. This is the cause of the problem. We should treat the cause rather than the symptom in my opinion.

I think the natural / go-to choice for newcomers to the ecosystem doesn’t actually need to be DFINITY. We can establish a better go-to option. One which is highly decentralised and competitively selective.

3 Likes

There is nothing ‘artificial’ about treating staked ICP as staked ICP - I believe it’s actually the other way around; this system artificially inflates the weight of voting power of “active neurons” by removing the voting power of “inactive” neurons.

Non-Voting Neurons do not receive voting rewards, which is their penalty for being an “inactive” neuron.

In this instance, we are shareholders, not appointed company executives. If a shareholder doesn’t vote, their vote isn’t void, they just abstain from contributing to either side of quorum. That doesn’t mean that the threshold to achieve 51% is lowered.

1 Like

I appreciate the work you do @Lorimer but I feel as though this is off topic as this doesn’t actually pertain to the problem; voting power of ‘inactive’ neurons being removed from the consensus for the benefit of ‘active’ neurons proportional weight of voting power.

While I agree that more Known Neurons and followee alternatives is a good thing, that wouldn’t serve to solve anything in this instance as the voting power would still be removed from consensus.

3 Likes

Thanks @Accumulating.icp, apologies for side-tracking the discussion. I think I might have misunderstood your concerns.

I see. I was focusing on the users who have now been prompted to reconfirm their following (many of which I assumed switched to DFINITY). But I now see that you’re primarily concerned about the VP which is being diminished as a result of those who are not reconfirming their following.

I can see why this may be a concern. Just to make sure I understand, am I right in thinking that your concerns would be addressed by raising the threshold for achieving a supermajority (which immediately executes a proposal)? I’m not saying that this should be done, but I’m just considering that this would be potentially equivalent to giving those ‘inactive’ neurons their VP back.

2 Likes

I can definitely understand this perspective - I believe that’s how most people have interpreted this. My concern isn’t with one party specifically, but they were a good example to highlight the effects of this proposal. I don’t think that DFINITY seen an increase in followees due to this (I may be wrong), but rather their weight of voting power increased due to the overall reduction.

I believe that consensus should always be the result of all staked ICP, whether that is a standard/super majority or rejection of a proposal.

3 Likes

I think this is an interesting point. I actually think there’s a flaw in PoS systems in general, unless it’s acknowledged and managed appropriately.

As a thought experiment, if 1 million people are careless with their 1 ICP stake (and not bother to manage and maintain who they’re following), is it okay that this puts another person’s 1 million ICP stake at risk (a person who is therefore diligent and thoughtful about who they follow and how it affects decentralisation)?

It’s a failing of proof of stake systems that many people with little to lose (and wilful abandon) have the potential to undermine the system if there are enough of them. Personally I think the VP is only securing the network if it’s being managed and maintained by those in charge of that VP.

3 Likes