Click hashing vulnerabilty?

[Eddie Kohler]

Hi Julian,

> I was just wondering if anyone has any comments on the vulnerability or
> otherwise of Click to attacks on the hashing algorithm as mentioned here:
> 
> http://www.cs.rice.edu/~scrosby/hash/
> 
> It specifically mentions the hashing of IPv4 & IPv6 values and mentions in
> its conclusion a recommendation for auditing network packet processing code.

Click doesn't use universal hashing. The elements which might be
'vulnerable' to a hash attack, or a more general address-faking attack,
include:

- several trace analysis elements, including AggregateIPFlows and
  AnonymizeIPAddr
- EtherSwitch (hashes Ethernet addresses, local attacks only)
- several Grid elements
- NAT elements:
  ICMPPingRewriter (NAT for pings)
  IPRewriter (NAT for TCP/UDP)
  TCPRewriter (NAT for TCP)
  AddressTranslator (NAT for IPv6->IPv4) and friends

  Note that, depending on how you set up your NAT, the elements might be
  vulnerable only to attacks that originate "inside" the NAT.

- RFC2507Comp (very old element, possibly doesn't work anyway :) )
- ARPQuerier

This list is probably complete. Only ARPQuerier seems worth worrying about.
While fixing these elements to use universal hashing is not on our priority
list, we welcome contributions from others!! (preferably #ifdefed)

Eddie