I have been following this list for quite some time and
have never heard this vulnerability mentioned.

If a server responded (either legitimately or not) to
all requests and sent the reply data out reeeaaalllyy
ssslllllooooowwwlllyyy, like 1 byte/sec, wouldn't that
seriously damage freenet's usefulness? If an evil
node that did this dedicated enough ports to doing
this, it could cause many, many other nodes to
just sit there and wait for the sluggish transfer to
complete. Splitting data into multiple parts would
make this even worse since there is a better chance
that the evil node with be contacted for a part of the
request.

The consequences:
- enough requests would be bogged down to
annoy a lot of people into not using freenet
- lots of nodes will use up their connection limit
- all nodes along a request chain will be affected

The solution:
If a supplying node is taking too long to transfer
data,  have the node randomly select another node
to try to get the data from and when it finds a better
one then it drops the slow-poke. A "minimum
acceptable transfer rate" in the config would allow
for this rate to be adjusted up as high bandwidth
becomes more wide spread.
This has the added benefit of biasing the whole
freenet towards faster nodes and culling the
traffic to slow ones (either maliciously slow or
just overloaded)

Mike


_______________________________________________
Freenet-dev mailing list
Freenet-dev at lists.sourceforge.net
http://lists.sourceforge.net/mailman/listinfo/freenet-dev

Reply via email to