#dusk $DUSK @Dusk $BTC 11 host queries shared the same wrapper. Eight had the same deserialization risk.
That made me look closer.
The contract controlled the bytes. The node was the one reading them. A malformed relative pointer could push the host toward an out of bounds read. At that point, the problem was no longer contained inside the contract.
What caught my attention was the fix.
Dusk did not patch eight queries separately. It fixed the shared boundary and validates the archived data before deserializing it.
That order matters.
Validate first. Deserialize second.
There was no useful user side workaround here. The boundary itself had to be fixed, and because the same wrapper served 11 queries, fixing that layer covered the whole group.
The test is simple. Feed malformed archived data into those queries. The node should reject it before the wrapper tries to interpret the bytes.
That made me look closer.
The contract controlled the bytes. The node was the one reading them. A malformed relative pointer could push the host toward an out of bounds read. At that point, the problem was no longer contained inside the contract.
What caught my attention was the fix.
Dusk did not patch eight queries separately. It fixed the shared boundary and validates the archived data before deserializing it.
That order matters.
Validate first. Deserialize second.
There was no useful user side workaround here. The boundary itself had to be fixed, and because the same wrapper served 11 queries, fixing that layer covered the whole group.
The test is simple. Feed malformed archived data into those queries. The node should reject it before the wrapper tries to interpret the bytes.