The latest release of Trisul Network Analytics [ release details ] has a really nifty self-monitoring feature. There is no need for outside tools like atop or nmon to check on Trisul’s performance any more.
To use the feature :
Login as admin
Select Tools > Perf Stats
The Perf stats dashboard is shown similar to the one below.
The modules shown in the Perf Stats dashboard are
Global Flush Time
Trisul is a streaming network analytics engine. Since there is a continuous stream of high volume data, one key requirement is that aggregated results are flushed out within a small amount of time. Global Flush-0, 1 are the two default backend threads of Trisul flushing out in parallel. You can have upto 8 backend threads in that case all 8 will be shown on the graph.
Key thing to look for : See if there is spikiness in the chart, all data should ideally be less than 20 seconds.
Shows Total and Trisul CPU Usage over time.
Lookout for : CPU being pegged. Ideally Trisul CPU usage must rise and fall with load. Unusual spikes in low traffic times have led us to find scans and DoS attacks as the Trisul TCP reassembly engine is stressed during these times.
Total and Trisul Memory Usage
Lookout for : Memory climbing , this is very very rare and could indicate a memory leak. Trisul has a lot of hi-water and lo-water marks to restrict memory usage. If your server has a lot of memory you could consider increasing these numbers so Trisul can make more judicious use of RAM.
Total and Trisul Usage. Shows the amount of disk used by the Trisul backend database.
Lookout for : If data climbs steadily and threatens to it the maximum, you may need to adjust the SlicePolicy to reduce the number of slices kept in the system. This way the DB size is kept smaller than the available disk space.
With this new feature, your Trisul monitors itself.
I stumbled across this post on “behindthefirewalls.com” blog about the recent PHP.com compromise titled “Extracting files from network traffic capture“. In that blog, the author has demonstrated file carving using Wireshark and other tools.
I’d like to introduce you to Unsniff Network Analyzer‘s nifty file extraction that addresses the following issues in the PCAP.
The EXEs are transferred as content type “text/html”
All files have to be written to disk before you can do a file * and pick out the EXEs
The latest version of Unsniff has two extremely useful features that can really speed up this process. Each User Object now has two new attributes
Magic String : We take the first 4 bytes of each content and create a human readable string
MD5 Hash : Each user object has a MD5 content hash
These are computed online as traffic is being processed. Once they are stored in an Unsniff Capture File Format (*.USNF) you can just access them instantaneously without reprocessing.
Magic number – pick out EXE transferred as text/html
As mentioned, the Barracuda PCAP drops EXE malware as text/html. In the screenshot below, If you notice User Object 11 – the Type column shows “HTML” but the Magic column shows “MZ90.00.” Thats a dead giveaway that the content isnt really HTML. Next, you can click on the corresponding MD5 column to start checking with VirusTotal etc.
Files like CSS/JS/HTML usually just have the first 4 bytes of text as the magic number. You can simply ignore them. The best part is this feature works for all files transferred – whether as EMAIL attachment, as FTP files, as Chat file transfers, etc, etc.
Saving all the Malware EXE files
The way you save the malware files is to simply “Ctrl-Click” and select all those who start with the EXE Magic number. Then Right Click and Save.
What gives Unsniff even more power is that the entire process above is scriptable in Ruby or VBScript. You never have to open a single GUI window. You can use the Unsniff Scripting API to automatically chew through PCAPS and continuously dump only the EXE files this way.
If you are into info sec, Unsniff Network Analyzer is a tool that you really need to have in your kit. Download for free today.
There was a question on the Wireshark Q&A Site. A user wanted to apply a filter to only show the last few packets of all TCP flows. This can be done manually in Wireshark quite easily but the user had hundreds of flows and was looking for an automatic way to do this.
Here is a quick post with code that demonstrates how you can automate this and other custom analysis using Unsniff Network Analyzer.
Unsniff exposes an object model to scriptland. This means that flows (a.k.a streams), packets, user objects, PDUs are all top level objects. All you have to do is to grab the flows collection, then for each flow save the last 6 packets. Here is the script.
'lastn-Dump last5packets of every stream intoaseparate PCAP file