Inspirated

As I promised in the comments section of previous post, I set out on the adventure of recompiling Bro for Lantiq routers. As a result of the exercise I have new-found respect for open-source package maintainers. Holy waffles if troubleshooting build errors in a large Autotools mess isn’t the most hemorrhage-inducing activity known to mankind.

Anyways, this time I’ve tried to keep track of the changes I’ve been making along the way. The full set of updated Makefiles and patches is maintained in the openwrt-bro repo. Also, the compiled ipk packages for Atheros and Lantiq routers are available on the release page.

Now that I have a reasonably updated Buildroot on my system and an organized set of patches, feel free to request an ipk package for your router. While I can’t guarantee that the clusterfuck of patches will compile smoothly for your platform, I’ll still give it a try.

After posting the Bro port for OpenWRT on my blog roughly two years ago, I didn’t realize some people were already actually using it on their routers.

I had created an updated version of the port which I hadn’t posted on the blog. Digging in my archived files I finally found it today along with its sources:

• openwrt-bro-2.3.1.tar.gz
• bro_2.3.1-1_ar71xx.ipk

Word of caution though, my notes indicate that one of the default scripts was leaking memory and I never got around to figuring out which one. The workaround was to launch Bro in barebone mode with -b switch, which would prevent loading of default scripts.

# cat test.bro

event bro_init()
{
	print "Hello World!";
}

event new_connection(c: connection)
{
	print "New connection created";
}

# bro test.bro

Hello World!

# bro -C -b -i br-lan test.bro

Hello World!
New connection created
New connection created

If someone has cycles to spend and figure out which default script is leaking memory we can update the package to address the bug.

The paper chronicling our adventures with Bro IDS on home routers just got published in the latest issue of SIGCOMM CCR. Here’re the details:

Title: Rapid and Scalable ISP Service Delivery through a Programmable MiddleBox

Abstract: With only access billing no longer ensuring profits, an ISP’s growth now relies on rolling out new and differentiated services. However, ISPs currently do not have a well-defined architecture for rapid, cost-effective, and scalable dissemination of new services. We present iSDF, a new SDN-enabled framework that can meet an ISP’s service delivery constraints concerning cost, scalability, deployment flexibility, and operational ease. We show that meeting these constraints necessitates an SDN philosophy for a centralized management plane, a decoupled (from data) control plane, and a programmable data plane at customer premises. We present an ISP service delivery framework (iSDF) that provides ISPs a domain-specific API for network function virtualization by leveraging a programmable middlebox built from commodity home-routers. It also includes an application server to disseminate, configure, and update ISP services. We develop and report results for three diverse ISP applications that demonstrate the practicality and flexibility of iSDF, namely distributed VPN (control plane decisions), pay-per-site (rapid deployment), and BitTorrent blocking (data plane processing).

Published in: ACM SIGCOMM Computer Communication Review (Volume 44 Issue 3, July 2014)

Combined with the paper in IEEE COMST about botnet detection that was published last year, this yields a grand-total of 2 publications more than I thought would ever bear my name. In any case, my former colleagues are continuing their excellent work on the project which can be tracked at the iSDF wiki-page.

I needed to block some flows on OpenWRT from the Bro IDS. One option was to install the recent module for expiring iptables rules which sounded like an overkill. After some tinkering around I landed on using bash and at to expire the firewall rules after timeouts (luckily the at daemon was available on OpenWRT which made my job easier).

There are three parts to the process:

The bash script

First, a script which:

1. Constructs and adds the iptables rule to the FORWARD chain.
2. Constructs the corresponding deletion rule.
3. Creates a temporary bash script, writes the rule to it, makes the new script self-deleting.
4. Schedules a launch of the temporary script with at command.

Here’s the script:

#!/bin/sh
 
if [ $# -le 5 ] ; then
  echo "usage: $0 proto src sport dst dport timeout"
  exit 1
fi
 
proto=$1
src=$2
sport=$3
dest=$4
dport=$5
timeout=$6
 
echo "  proto: $1"
echo "    src: $2"
echo "  sport: $3"
echo "   dest: $4"
echo "  dport: $5"
echo "timeout: $6"
 
rule=""
 
if [ "$proto" != "any" ]; then
  rule="$rule --protocol $proto"
fi
 
if [ "$src" != "0.0.0.0" ]; then
  rule="$rule --source $src"
fi
 
if [ "$sport" != "0" ]; then
  rule="$rule --sport $sport"
fi
 
if [ "$dest" != "0.0.0.0" ]; then
  rule="$rule --destination $dest"
fi
 
if [ "$dport" != "0" ]; then
  rule="$rule --dport $dport"
fi
 
rule="$rule -j DROP"
 
echo "rule: $rule"
 
addcmd="iptables -I FORWARD $rule"
delcmd="iptables -D FORWARD $rule"
 
delscript=`mktemp`
echo "delscript: $delscript"
 
echo "#!/bin/sh" >>$delscript
echo $delcmd >>$delscript
echo "rm \"${delscript}\"" >>$delscript
chmod 755 $delscript
 
echo "adding iptable rule:"
echo $addcmd
`$addcmd`
 
atcmd="at -M -f $delscript now + $timeout min"
echo "creating at job for deletion:"
echo $atcmd
`$atcmd`

Given below is an example run. First, let’s print the default FORWARD chain:

# iptables -nL FORWARD

Chain FORWARD (policy ACCEPT)
target     prot opt source               destination         
ACCEPT     all  --  anywhere             10.42.0.0/24         state RELATED,ESTABLISHED
ACCEPT     all  --  10.42.0.0/24         anywhere            
ACCEPT     all  --  anywhere             anywhere            
REJECT     all  --  anywhere             anywhere             reject-with icmp-port-unreachable
REJECT     all  --  anywhere             anywhere             reject-with icmp-port-unreachable
REJECT     all  --  anywhere             anywhere             reject-with icmp-host-prohibited

Block a flow for 2 minutes:

# sh blockflow.sh tcp 50.50.50.50 50 60.60.60.60 60 2

  proto: tcp
    src: 50.50.50.50
  sport: 50
   dest: 60.60.60.60
  dport: 60
timeout: 2
rule:  --protocol tcp --source 50.50.50.50 --sport 50 --destination 60.60.60.60 --dport 60 -j DROP
delscript: /tmp/tmp.SAREJvtsK0
adding iptable rule:
iptables -I FORWARD --protocol tcp --source 50.50.50.50 --sport 50 --destination 60.60.60.60 --dport 60 -j DROP
creating at job for deletion:
at -M -f /tmp/tmp.SAREJvtsK0 now + 2 min
job 79 at Sun Jun 30 14:37:00 2013

Let’s check if the new rule was added:

# iptables -nL FORWARD

Chain FORWARD (policy ACCEPT)
target     prot opt source               destination         
DROP       tcp  --  50.50.50.50          60.60.60.60          tcp spt:50 dpt:60
ACCEPT     all  --  anywhere             10.42.0.0/24         state RELATED,ESTABLISHED
ACCEPT     all  --  10.42.0.0/24         anywhere            
ACCEPT     all  --  anywhere             anywhere            
REJECT     all  --  anywhere             anywhere             reject-with icmp-port-unreachable
REJECT     all  --  anywhere             anywhere             reject-with icmp-port-unreachable
REJECT     all  --  anywhere             anywhere             reject-with icmp-host-prohibited

After 2 minutes, the temporary bash script shall remove the rule and then delete itself. To confirm:

# iptables -nL FORWARD

Chain FORWARD (policy ACCEPT)
target     prot opt source               destination         
ACCEPT     all  --  anywhere             10.42.0.0/24         state RELATED,ESTABLISHED
ACCEPT     all  --  10.42.0.0/24         anywhere            
ACCEPT     all  --  anywhere             anywhere            
REJECT     all  --  anywhere             anywhere             reject-with icmp-port-unreachable
REJECT     all  --  anywhere             anywhere             reject-with icmp-port-unreachable
REJECT     all  --  anywhere             anywhere             reject-with icmp-host-prohibited

The Bro module

A simple module which exports just one function, i.e., BlockFlow::block which takes a conn_id and a count and calls the bash script with appropriate parameters:

module BlockFlow;
 
export {
  global block: function(id: conn_id, t: count);
}
 
function block(id: conn_id, t: count)
{
  print fmt("blocking %s:%d -> %s:%d for %d minutes", id$orig_h, id$orig_p, id$resp_h, id$resp_p, t);
 
  local protocol = get_port_transport_proto(id$resp_p);
  print fmt("protocol is: %s", protocol);
 
  local cmd: string = fmt("sh blockflow.sh %s %s %d %s %d %d", protocol
                                                             , id$orig_h, id$orig_p
                                                             , id$resp_h, id$resp_p, t);
  print fmt("executing: %s", cmd);
  system(cmd);
}

Bro module usage

And finally, using the module from a Bro script:

@load ./blockflow
 
event bro_init()
  {
    local id: conn_id;
    id$orig_h = 10.10.10.10;
    id$orig_p = 10/tcp;
    id$resp_h = 20.20.20.20;
    id$resp_p = 20/tcp;
    BlockFlow::block(id, 2);
  }

And the flow will be blocked for 2 minutes. Unfortunately, due to the way at command works the granularity of timeouts is limited to minutes. If you really want to block flows for only a few seconds a quick solution would be to use sleep in place of at before expiring the rule.

While I was at SysNet, we had been working on a project we called “Shrimp” — Software-defined Home Router Intelligent Monitoring Point. The goal of the project was to provide a framework for easy programmatic access to network monitoring on low-cost, commodity, home router devices. One of the requirements was to have an IDS on the home routers for which we chose Bro — the leading framework for semantic analysis of network traffic.

The OpenWRT OS was chosen as the target platform. Its SDK contained a cross-compile toolchain for CMake projects. However, during the compilation Bro tried to run the binpac and bifcl executables for processing intermediate files. The executables refused to run on the build platform if the target platform architecture was different (mostly the case, e.g., we were building on x86-64 and target was arm).

The (not-so-pretty ™) workaround we used was to build Bro twice. Once for the host, and once for the target. The CMake files were then patched to first generate binpac and bifcl binaries if they weren’t provided and then use the provided binaries if they were defined at make time. The first compile generated the binaries on x86-64 and the second compile (for arm) used the earlier binaries to process the bif files.

The Makefile and patches are available in this tarball: openwrt-bro.tar.gz, while the compiled ipk package is also available for installation. Here is a test execution of Bro on OpenWRT:

# bro –v

bro version 2.0

# cat test.bro

event bro_init()
{
	print "Hello World!";
}

event new_connection(c: connection)
{
	print "New connection created";
}

# bro test.bro

Hello World!

# bro -i br-lan test.bro

Hello World!
New connection created
New connection created

# ls

conn.log           notice_policy.log  reporter.log       weird.log
dns.log            packet_filter.log  test.bro

A heap of thanks to Zaafar for dealing with my messy code and providing the links to hosted files :) !