-
Notifications
You must be signed in to change notification settings - Fork 2.3k
Prune unconnected nodes from the channel graph #138
Copy link
Copy link
Closed
Labels
databaseRelated to the database/storage of LNDRelated to the database/storage of LNDdos/hardeningRelated to the resilience of LND against denial of service or other related attacksRelated to the resilience of LND against denial of service or other related attacksintermediateIssues suitable for developers moderately familiar with the codebase and LNIssues suitable for developers moderately familiar with the codebase and LNrouting
Milestone
Description
Metadata
Metadata
Assignees
Labels
databaseRelated to the database/storage of LNDRelated to the database/storage of LNDdos/hardeningRelated to the resilience of LND against denial of service or other related attacksRelated to the resilience of LND against denial of service or other related attacksintermediateIssues suitable for developers moderately familiar with the codebase and LNIssues suitable for developers moderately familiar with the codebase and LNrouting
Currently we accept any and all
NodeAnnouncementmessages received. Instead we should:Limit the number of unconnected nodes we accept and write to the database.
Remove
LinkNodes from the database, once we have closed out all previous channels with them.The lack of an implementation of #1 opens us up to a disk flooding attacking whereby a node connects and send us a bunch of "fake" nodes with pubkeys that it generated on the fly. This attack is very cheap compared to channel creation as it is very cheap to generate new public keys, generate the signature then send us a bogus
NodeAnnouncementmessage.The lack of an implementation of #2 atm, results in the daemon attempting to establish persistent connections to nodes that it doesn't have any channels with.