Apr 162014

After having fun with Suricata’s new eve/json logging format and the Logstash/Elastic Search/Kibana combination (see this and this), I wanted to get my Snort events into Elastic Search as well.  Using my idstools python library I wrote u2json, a tool that will process a unified2 spool directory (much like barnyard) and convert the events to Suricata-style JSON.

Usage is relatively simple, assuming Snort is logging to /var/log/snort, the following command line should do:

  idstools-u2json -c /etc/snort/snort.conf 
--directory /var/log/snort
--prefix unified2.log
--follow --bookmark
--output /var/log/snort/alerts.json

As the output is in the same format as Suricata’s you can refer to this guide for the Logstash setup.

One extra step I did was use Logstash to add an “engine” field to each entry.  This can be accomplished by adapting the following Logstash configuration:

input {
file {
path => ["/var/log/suricata/eve.json"]
codec => json
type => "suricata-json"
file {
path => ["/var/log/snort/alerts.json"]
codec => json
type => "snort-json"

filter {
if [type] == "suricata-json" {
mutate {
add_field => {
"engine" => "suricata"

if [type] == "snort-json" {
mutate {
add_field => {
"engine" => "snort"

Checkout out the documentation for information.

Jul 192013

I recently consolidated my Python code bits for dealing with Snort and Suricata unified2 log files into a project called idstools. While I’ll be adding more than just unified2 reading support, that is about it for now. 

While it can be installed with pip (pip install idstools), if you just want to play around with it I suggest cloning the repo (git clone https://github.com/jasonish/idstools.py). You can then use the REPL or write test scripts from within the idstools.py directory without having to install the library (yeah, basic stuff for Python developers).

idstools does come with a few example programs that demonstrate unified2 file reading, namely, u2fast.py, u2tail.py and u2spewfoo.py (a simple clone of the Snort provided u2spewfoo).

Basic Unified2 File Reading

from idstools import unified2

reader = unified2.FileEventReader("tests/merged.log")
for event in reader:
print("Event:n%s" % str(event))

This few lines of code will iterate through each record in the specified unified2 log files, aggregate the records into events and return each event as a dict.

If straight up record is reading is more what you are after then check out unified2.FileRecordReader, or the lower level unified2.read_record function.

Each event is represented as a dict containing the fields of a unified2 event record, with the associated packets represented as a list in event["packets"] and extra data records represented as a list in event["extra-data"].

Resolving Event Message and Classification Names

To make event reading just a little more useful, code to map signature and classifications IDs to descriptions is provided.
from idstools import maps

# Create and populate the signature message map.
sigmap = maps.MsgMap()

# Get the description for 1:498.
print("Message for 1:498: %s" % (sigmap.get(1, 498).msg))

# Create and populate the classification map.
classmap = maps.ClassificationMap()
print("The description for classification id 9 is %s, with priority %d." % (
classmap.get(9).description, classmap.get(9).priority))
The example program u2fast.py is a complete example of reading events from one or more files, resolving event descriptions and classification names and printing the event in a “fast” like style.

Spool Reading

idstools also contains a spool reader for processing a unified2 spool directory as commonly used by Snort and Suricata.  It supports bookmarking, deleting files, and open and close hooks which can be used to implement custom archiving.
from idstools import spool

def my_open_hook(reader, filename):
print("File %s has been opened." % (filename))

def my_close_hook(reader, filename):
print("File %s has been closed." % (filename))

reader = spool.Unified2EventSpoolReader(
"/var/log/snort", "merged.log", delete_on_close=False,

for event in reader:
print("Read event with generator-id %d, signature-id %d." % (
event["signature-id"], event["generator-id"]))
To see a more complete directory spool process, check out the u2tail.py example program.

To learn more checkout idstools over at GitHub, PyPI, or the work-in-progress documentation on Read the Docs.

Dec 182012

Sometimes the best way to try out a new framework or language is to apply it to a domain you already know very well, even if it does happen to reinvent the wheel.  Tornado and Twitter Bootstrap are two such frameworks I’ve been meaning to play with for a while now. The result is Dumpy, a web front-end to pcap spool files as created by tcpdump, daemonlogger, or netsniff-ng with a very simple configuration and user interface:

Requirements are minimal, Python 2.6 (so it will run on CentOS 6 with little hassle), Tornado and py-bcrypt which are both trivially installed with pip. It provides its own http server with SSL support, and does not require a database.

Usage is also simple.  Simply enter a pcap filter, or paste in a Snort or Suricata event in “fast” format, choose  start and end times (or simply offsets) and hit download.

If interested, start a pcap spool (ie: sudo tcpdump -i eth0 -C 1000 -W10 -G 3600 -w /tmp/eth0.log.%Y%m%d.) then check out Dumpy over here https://bitbucket.org/ish/dumpy.

Dec 132012

As mentioned in my previous post on using embedded Jetty with SpringMVC, I was going to look at simplifying the application by using Thymeleaf instead of JSPs as a view technology.

Well, it turns out that it is much more straightforward.

First we can remove the JSPC plugin from our pom.xml. Second we can completely remove the web.xml file.

Bootstrapping Jetty is much simpler now. We do not have to hook into the Jetty startup process with a lifecycle listener, instead we can directly create the dispatcher servlet and add it to a ServletContextHandler like we would any other servlet:

From JettyConfiguration.java:

public ServletHolder dispatcherServlet() {
AnnotationConfigWebApplicationContext ctx =
new AnnotationConfigWebApplicationContext();
DispatcherServlet servlet = new DispatcherServlet(ctx);
ServletHolder holder = new ServletHolder("dispatcher-servlet", servlet);
return holder;

public ServletContextHandler servletContext() throws IOException {
ServletContextHandler handler = new ServletContextHandler();
new ClassPathResource("webapp").getURI().toString());
handler.addServlet(AdminServlet.class, "/metrics/*");
handler.addServlet(dispatcherServlet(), "/");
return handler;

@Bean(initMethod = "start", destroyMethod = "stop")
public Server jettyServer() throws IOException {
Server server = new Server();
return server;

This is much more straightforward, everything wired up with Spring without any lifecycle callback hooks.

The setup for Spring to render the Thymeleaf views, is a little more complex, but not fussy. What we do is replace:

public InternalResourceViewResolver configureInternalResourceViewResolver() {
InternalResourceViewResolver resolver =
new InternalResourceViewResolver();
return resolver;


public ServletContextTemplateResolver thymeleafTemplateResolver() {
ServletContextTemplateResolver resolver =
new ServletContextTemplateResolver();
return resolver;

public SpringTemplateEngine thymeleafTemplateEngine() {
SpringTemplateEngine engine = new SpringTemplateEngine();
return engine;

public ThymeleafViewResolver thymeleafViewResolver() {
ThymeleafViewResolver resolver = new ThymeleafViewResolver();
return resolver;

in MvcConfiguration.java.

Our templates now look a lot more like plain HTML, and will render better when loaded directly in a browser:

<!DOCTYPE html>
<html xmlns:th="http://www.thymeleaf.org">
<h1>Hello World!</h1>

<p>Server time: <span th:text="${serverTime}"></span></p>

<p>Here are some items:</p>
<li th:each="item : ${someItems}" th:text="${item}"></li>

<p>Do we have a message from the dummy service:</p>

<div th:if="${dummyService == null}">
<p>No, dummy service is null.</p>
<div th:if="${dummyService != null}">
<p>Yes: <span th:text="${dummyService.getMessage()}"></span></p>

<p><a href="resources/static.txt">A static file.</a></p>

<p><a href="metrics">Yammer Metrics</a></p>


With the build process being simple (no dependence on maven plugins) we can switch to a much less verbose Gradle build file:

apply plugin:'java'
apply plugin:'application'

version = '0.0.1-SNAPSHOT'

mainClassName = "ca.unx.template.Main"
applicationName = "jetty-springmvc-thymeleaf-template"

repositories {

dependencies {
compile("org.springframework:spring-webmvc:3.1.3.RELEASE") {
// Commons-logging excluded in favour of SLF4j.
exclude module: 'commons-logging'

compile "cglib:cglib:2.2.2"
compile "org.thymeleaf:thymeleaf-spring3:2.0.14"
compile "org.eclipse.jetty:jetty-webapp:8.1.8.v20121106"
compile "com.yammer.metrics:metrics-servlet:2.2.0"

/* Logging. */
def slf4jVersion = "1.7.1"
compile "ch.qos.logback:logback-classic:1.0.9"
compile "org.slf4j:slf4j-api:$slf4jVersion"
compile "org.slf4j:jcl-over-slf4j:$slf4jVersion"
compile "org.codehaus.groovy:groovy:1.8.6"

A complete template project can be found here. Note that this links to tag which is the state of the code at the time of this writing.

Dec 072012

As an accidental Java developer I’ve never been comfortable deploying applications into a container, especially when the web interface is secondary to the primary purpose of the application. Instead I prefer to programatically create and manage the web interface rather than have it manage me. Currently SpringMVC is my Java web framework of choice (due to company convention more than anything else) and it is built around the idea of being managed by a container such as Jetty or Tomcat.

While there is no shortage of existing posts on using SpringMVC with embedded Jetty, they fail for me due to the following reasons:

  • They embed Jetty just enough to bootstrap something that looks like a classic Java web application – not what I want!
  • They don’t address JSPs – the default view technology used by SpringMVC.

Compile Those JSPs

By compiling the JSPs prior to deployment we can save ourselves from the hassle of classpath issues, especially relating to tag libs.


This plugin will compile all the JSPs found under src/main/resources/webapp and modify the existing web.xml to direct requests for the JSPs to the pre-compiled versions. This avoids compiling them at runtime and speeds up the time it takes to respond to the first request for a JSP.

You may also notice that the JSPs are under src/main/resources/webapp instead of the more common src/main/webapp. This is to avoid extra lines in the pom.xml to pull in webapp as a resource, as we are packaging as a jar.

Bootstrapping Jetty

I use Spring annotations to configure Jetty and import its @Configuration class into my root context configuration class. In order for the SpringMVC dispatcher servlet to access beans created in the root context, this class needs to be application context aware:

public class JettyConfiguration implements ApplicationContextAware {
private ApplicationContext applicationContext;

public void setApplicationContext(ApplicationContext applicationContext)
throws BeansException {
this.applicationContext = applicationContext;

this is used as the parent context for the new context that will be created for the SpringMVC dispatcher servlet:

public ServletHolder dispatcherServlet() {
AnnotationConfigWebApplicationContext ctx =
new AnnotationConfigWebApplicationContext();
DispatcherServlet servlet = new DispatcherServlet(ctx);
ServletHolder holder = new ServletHolder("dispatcher-servlet",
return holder;

There are still a few more tricks required to get the dispatcher servlet registered without defining it in your web.xml. We must also create a Jetty web application context, this creates the JSP servlet as well as a default servlet as would be created during the instantiation of a classic war based application:

public WebAppContext jettyWebAppContext() throws IOException {
WebAppContext ctx = new WebAppContext();
ctx.setWar(new ClassPathResource("webapp").getURI().toString());

/* We can add the Metrics servlet right away. */
ctx.addServlet(AdminServlet.class, "/metrics/*");

return ctx;

Notice that I’ve added the metrics servlet, but have yet to add the dispatcher servlet. For some reason, adding the dispatcher servlet here causes JSP views to fail. One option is to fall back to defining the dispatcher servlet in your web.xml (which may be the cleanest option), or you can register the dispatcher servlet in the Jetty lifeCycleStarted callback which we’ll do here:

public LifeCycle.Listener lifeCycleStartedListener() {
return new AbstractLifeCycle.AbstractLifeCycleListener() {
public void lifeCycleStarted(LifeCycle event) {
try {
ServletHolder dispatcherServlet = dispatcherServlet();
.addServletWithMapping(dispatcherServlet, "/");
} catch (Exception e) {
"Failed to start Spring MVC dispatcher servlet", e);

@Bean(initMethod = "start", destroyMethod = "stop")
public Server jettyServer() throws IOException {
Server server = new Server();

/* Add a life cycle listener so we can register the SpringMVC dispatcher
* servlet after the web application context has been started. */

return server;

In summary, this is just another variation of James Ward’s post on Containerless Spring MVC. The complete code for this project (which I use as a template) can be at https://github.com/jasonish/jetty-springmvc-jsp-template. The code as referenced in this post can be found in the 20121207 tag.

Next – As I’m not really a fan of JSPs, or the extra hoops required to make this work right I’ll probably look at updating the template with Thymeleaf support, which should greatly simplify things.

Apr 132012

I’ve always maintained more or less up to date Snort RPMs for RHEL for personal use and have recently added Suricata. As they may be useful for others I have cleaned them up a little and made a YUM repository for EL6 i386 and x86_64. See the http://nsm-rpms.unx.ca/ for more info.

A few things to note:

  • These RPMs use a prefix of /opt/nsm to prevent conflict with similar RPMs you may have installed, its a little bit out of the norm for RPMs and I’m open to comments…
  • Snort and Suricata packages will never be automatically upgraded as often upgrading to a new version requires some administration work such as updating your configuration files. To facilitate this the packages have their version as part of the name and “-latest” pseudo-packages are provided which will always install the latest RPM but you will have to “snort-select” or “suricata-select” the new version to make it active. I’ll probably have to add some more detailed documentation about this on the wiki.
As I’m also a regular Fedora user I’ll probably add Fedora builds at some point as its little effort to me. If Fedora builds would be useful to you please let me know and I may do it sooner than later.
Apr 222010

Recent issues with Linux on a laptop has given me the desire to try working on a Mac again. Even though I have a pretty well supported laptop (Lenovo x61), its the little issues that have begun to bother me which include not great sleep/suspend support, wireless and audio issues and an almost never ending stream of updates. Yes, you Linux people may say it works fine, but if you’ve ever used a Mac before you’ll know its not as good as a could be, by far.

I’ve also found that its really nice to have the real Microsoft Office handy when you swap documents with other non-Linux using people (yeah, OpenOffice doesn’t quite cut it when taking turns editing documents) which meant often having a Windows VM running.

The other factor that helps the Mac be a viable option is the majority of the software I develop already builds and installs on a Mac, or can do so with very minor modifications. The stuff that doesn’t is so tied to a specific environment that it really requires a virtual machine even when running Linux as the host operating system. This really makes the Mac sound like an ideal system.

So yesterday while heading out just before dinner to grab some hamburger buns I walked by the local Mac retailer (hey, its about 20 steps from Safeway) and grabbed a high-res 15″ MacBook with the Core i5 2.53ghz processor. They weren’t able to upgrade the ram to 8GB on the spot but I decided I couldn’t wait.

Aesthetically its a beautiful machine. The display is about the best computer display I’ve ever seen. The 3 primary pieces of software I work on compile out of the box. Its also nice to have Microsoft Office running natively. The thing that drove me away from the Mac before was some of the UI behaviours, like Cmd-Tab switching applications, not just windows of applications. But I’m going to give this a true go and may update this blog now and then with issues that I run into as a developer, developing network-level apps primarily for Linux, making the shift to Mac as a development platform.

Oct 072009

Probably the most lacking feature in most multi-screen setups is the ability to move a window from one screen to another with a keyboard shortcut. I do this to bring the window I’m currently working on into my centre monitor so its directly in front of me. There are 2 ways to go about this. The first is moving a window to a specific display and the second is to move (shift) the window to the left or right display (which would just be back and forth in a dual monitor setup).

For the first case you could define the following function:

# This function will move a window to the X screen provided in the

# first argument. If the window is in the maximized it will be taken

# out of the maximized state before moving. If we don’t do this it

# will be moved back to the original screen when the maximized state

# is toggled.


# If you do not want the pointer to be moved to the windows new

# location then comment out the line containing WarpToWindow.

DestroyFunc MoveWindowToScreen

AddToFunc MoveWindowToScreen

+ I Current (Maximized) Maximize

+ I SetEnv PWX $[pointer.wx]

+ I Setenv PWY $[pointer.wy]

+ I MoveToScreen $[0]

+ I WindowId $[w.id] WarpToWindow $[PWX]p $[PWY]p

And then add the following to your window operations menu:

+ “&1 Move to Screen 1″ MoveWindowToScreen 2

+ “&2 Move to Screen 2″ MoveWindowToScreen 0

+ “&3 Move to Screen 3″ MoveWindowToScreen 1

Note that the argument you provide to the MoveWindowToScreen function takes the X display number which may not directly map to the order of the screens on your desk.

The second method is to move the screen to the left or the right. Or in the case of a dual monitor setup you could just continually move right to have the window flip/flop between your 2 screens.

First download the following file and save it to ~/.fvwm/ShiftToScreen.py: ShiftToScreen.py

You will also need to update variables screenWidth and screens in ShiftToScreen.py as they are currently setup for my configuration of 3 screens with a width of 1920 pixels each.

Then define the following function:

# This function will move a window to the left or right screen

# depending on the first argument which should be on of “left” or

# “right”. It depends on the auxiliary python script

# ShiftToScreen.py.

DestroyFunc ShiftToScreen

AddToFunc ShiftToScreen

+ I Current (Maximized) Maximize

+ I SetEnv PWX $[pointer.wx]

+ I Setenv PWY $[pointer.wy]

+ I PipeRead “python $./ShiftToScreen.py $[0] $[w.x]”

+ I WindowId $[w.id] WarpToWindow $[PWX]p $[PWY]p

You could then add the following to your windows operations menu:

+ “&Left Screen” ShiftToScreen left

+ “&Right Screen” ShiftToScreen right

or define some keyboard shortcuts. For example I have bound Shift-Alt-Left and Shift-Alt-Right to shift the window to the left or right screen:

Key Left WTSF12468 SM ShiftToScreen left

Key Right WTSF12468 SM ShiftToScreen right

If you are only working with 2 screens you will only need to define one shortcut to get flip/flop like behaviour as shifting left beyond the leftmost screen will wrap the window around to the right most screen.


Sep 222009

Multi-monitor configurations are great but I think they leave out some usability aspects that could really make life easier for the keyboard user. When I was just using dual displays I really wanted keyboard shortcuts to do the following:

* Move a window from one screen to the other.
* Swap what is seen on each display.

When I moved to 3 monitors my needs changed a little.

* Shift a window to the screen to the left or the right, rotating around.
* Swap a display with the one to the left or the right.
* Shift all displays to the left or the right.

This would make it extremely fast to move the window you are currently working on to your primary display, then push it back to where it came from. Or swap the contents of your right screen with your centre screen if you need to shift your work focus for a while.

As far as I know this is not possible with Windows, Mac or most Linux configurations. However, I believe using FVWM as your Linux (ok, *nix) window manager will make this all possible, at least my experiments with dual screens have been positive. I’ll update this blog with various functions that should make all of the above achievable.