I found one AEGIS detail more revealing than the headline audit count.
Dusk Network did not enforce the Phoenix fee fix only in the mempool.
The new consistency check verifies that gas limit multiplied by gas price matches the proven maximum fee. Dusk applies it when a transaction enters the mempool and again during VM execution.
The duplication is deliberate.
A mempool is an admission layer. It can reject malformed transactions before they consume block space. But a block proposer is not forced to use transactions that passed another node's admission path. If the check existed only there, a malicious proposer could bypass the filter and push invalid semantics toward execution.
The VM check moves the invariant to the point where every accepted block must respect it.
I like this because it separates efficiency from consensus safety. Early rejection protects nodes and bandwidth. Execution-time rejection protects protocol state. The same condition serves 2 purposes, and removing either one weakens a boundary.
This pattern reaches beyond one fee bug. Dusk is building infrastructure for regulated assets, where applications may precheck eligibility, balances, limits, and settlement conditions before submission. Those checks improve user experience. They cannot become the only enforcement if a direct contract call or hostile proposer can avoid them.
The authoritative rule must survive the least cooperative path.
I would watch Dusk applications for the same discipline. Simulate a transfer before charging the user, then enforce the rule again inside the contract. Validate a credential at the interface, then verify it where asset state changes. Use the sequencer for ordering, but do not confuse sequencing policy with final settlement.
AEGIS fixed one invariant at 2 layers. That is not redundant engineering. It is an admission that honest infrastructure cannot be the security assumption.
For financial markets on Dusk, convenience checks should fail early. State-safety checks must fail everywhere.
@Dusk_Foundation $DUSK #dusk
$ACE $CYS
Dusk Network did not enforce the Phoenix fee fix only in the mempool.
The new consistency check verifies that gas limit multiplied by gas price matches the proven maximum fee. Dusk applies it when a transaction enters the mempool and again during VM execution.
The duplication is deliberate.
A mempool is an admission layer. It can reject malformed transactions before they consume block space. But a block proposer is not forced to use transactions that passed another node's admission path. If the check existed only there, a malicious proposer could bypass the filter and push invalid semantics toward execution.
The VM check moves the invariant to the point where every accepted block must respect it.
I like this because it separates efficiency from consensus safety. Early rejection protects nodes and bandwidth. Execution-time rejection protects protocol state. The same condition serves 2 purposes, and removing either one weakens a boundary.
This pattern reaches beyond one fee bug. Dusk is building infrastructure for regulated assets, where applications may precheck eligibility, balances, limits, and settlement conditions before submission. Those checks improve user experience. They cannot become the only enforcement if a direct contract call or hostile proposer can avoid them.
The authoritative rule must survive the least cooperative path.
I would watch Dusk applications for the same discipline. Simulate a transfer before charging the user, then enforce the rule again inside the contract. Validate a credential at the interface, then verify it where asset state changes. Use the sequencer for ordering, but do not confuse sequencing policy with final settlement.
AEGIS fixed one invariant at 2 layers. That is not redundant engineering. It is an admission that honest infrastructure cannot be the security assumption.
For financial markets on Dusk, convenience checks should fail early. State-safety checks must fail everywhere.
@Dusk_Foundation $DUSK #dusk
$ACE $CYS