Skip to content
This repository has been archived by the owner on Nov 10, 2018. It is now read-only.

Latest commit

 

History

History
59 lines (43 loc) · 3.83 KB

README.md

File metadata and controls

59 lines (43 loc) · 3.83 KB
Raw

snitchd

snitchd is a loadable kernel module (LKM) for Linux. It periodically scans all of physical memory looking for various types of sensitive data:

  • Credit card numbers (PANs) -- based on a simple regex and a Luhn checksum validation.
  • Social Security numbers -- based on a simple regex and a few validations.
  • First and last names -- based on a list of common US names taken from US census data.
  • Email addresses -- based on a simple regex.
  • Routing numbers -- based on a list of valid numbers from the Federal Reserve. The routing portion of these numbers isn't sensitive by itself, but these are often paired with a full bank account number.

When it finds data that looks like one of these types, snitchd counts it and reports via Graphite. This data can be used by other tools to detect when data is on a system where it doesn't belong.

Because the data type detection is heuristic, it's likely to detect several false positives on most systems. Detection and alerting logic based on the snitchd metrics should take this into account and do some basic filtering.

Caveats

This code is not necessarily production ready. It is highly privileged code written in an unsafe language and exposed to a large amount of attacker-controlled input. Use with caution.

Threat Model

snitchd is not meant to detect malicious copies of data. It is meant to help detect accidental leakage of data onto systems which are not meant to handle that type of data.

Example Use Cases

These are some examples of security policies that could be enforced with the help of snitchd:

  • A development/test cluster of machines should not store or process any sensitive customer data.

  • Sensitive customer data should be encrypted at the application level before being stored in a database, so the database host should not contain any sensitive data.

  • Data sent through a load balancer should be encrypted end-to-end between the client and the upstream server, so buffered data on the load balancer host should not contain any sensitive data.

Building

snitchd is built using Ragel, a tool for writing efficient and safe pattern matching code using state machines. On Ubuntu 14.04, you can get all the dependencies using sudo apt-get install build-essential ragel. The code should then be buildable with make.

Usage

snitchd depends on the keyutils package which provides helpers for resolving DNS names from kernel modules (sudo apt-get install keyutils).

It can then be loaded using insmod:

sudo insmod snitchd-3.13.0-54-generic.ko interval=900 graphite_host=127.0.0.1 graphite_port=8125 graphite_prefix=snitchd.myhost
  • interval: how often to scan in seconds (default: 15 minutes)
  • graphite_host: Graphite hostname or IP (default: 127.0.0.1)
  • graphite_port: TCP port for Graphite (default: 2003).
  • graphite_prefix: prefix for values sent to Graphite (required)
  • pan_prefix: count possible PANs only if they begin with this value (default: no filter)

Output Format

Output is in Graphite format, prefixed with the graphite_prefix:

snitchd.myhost.pans <number PAN-like values> <timestamp>
snitchd.myhost.ssns <number of SSN-like values> <timestamp>
snitchd.myhost.routing_numbers <number of routing number-like values> <timestamp>
snitchd.myhost.names <number of common names> <timestamp>
snitchd.myhost.emails <number of email-like values> <timestamp>

Notice

Census data is public domain / CC0 1.0 from https://deron.meranda.us/data/.

We can't distribute the Federal Reserve ACH routing number database (FedACHdir.txt), but the included Makefile should prompt you to download it.