How mitmproxy works | Hacker News Reader

How mitmproxy works

corte.si

120 pointsveb11 comments

This is a great, informative post; I'd just point out that generating on-the-fly certs is a basic functionality of all the web app testing tools, Burp Suite being the most famous.

If you haven't looked at Burp Suite and, unlike Aldo Cortesi, can stomach running a Java program (I don't like Java either, but...), Burp is both the industry standard tool used by appsec testing for web work and pretty inexpensive for what it does.

I run plenty of Java every day without breaking out in hives. I don't particularly enjoy _writing_ Java, though, and the most natural way to extend the Java interceptors is to use Java (Jython notwithstanding). I could have been clearer about this in my post.

I should also add - Burp is a fine tool, but I would hope there's room for more than a single "industry standard" in this area. Mitmproxy has some comparative strengths, is completely open, and is improving rapidly. I'd hate for people to be discouraged from trying it out and helping to make it better.

Well, mitmproxy isn't really just a Burp replacement. I've used both Burp and libmproxy for app sec reasons, but when I had strange reasons to write my own caching server for my crawler infrastructure, I dropped libmproxy in, wrote out some controller logic on top, and it just worked. I've been laughing every step to the bank ever since.

(bias: submitted a few patches to mitmproxy.git)

From the documentation, the mitmdump utility is likened to tcpdump. Can mitmdump work without being configured as a proxy/router, ie: using libpcap/tcpdump?

I guess it wouldn't really be mitm in this regard, but manipulating the traffic is not always needed.

Currently I use httpry to watch http traffic without interfering with its flow, but httpry is fairly limited in its functionality.

By the way, very impressive piece of software. Its nice to have more quality open tools like this.

[edit: I actually misread your question. We don't at the moment have reassembly of HTTP flows from pcaps in base, but there's an external project that does this. As another comment noted, transparent mode is probably still your best bet. https://github.com/cjneasbi/mitmextract]

Yes, it's very similar to tcpdump in this respect. You can write flows to disk like this:

   mitmdump -w outfile

And then read them back:

   mitmdump -r outfile

All the standard options for modifying, replaying and filtering are available when you read saved flows. So you could do this:

   mitmdump -r outfile -s foo.py -w newfile

Which will read flows from file, run the script foo.py over each flow as its read (which can then modify them arbitrarily), and then write the result to newfile.

Transparent mode is great but invasive, and mitmextract is fine for non real-time analysis.

Sometimes you might want to analyze traffic on production systems and changing routes is not an option.

What I would love to see is a way of doing a non-invasive, realtime processing of http traffic, eg when a specific post request is made, trigger some job.

That would be nice indeed. Mitmextract uses libnids, which can do stream reassembly on the fly. It's not such a big step going from reassembling a pcap file to doing the same thing on network traffic.

Look up the transparent mode

Great article really! I develop mobile application and you don't know how your software is indispensable for me :) i use it quite everyday really appreciated

mitmproxy = Malcolm in the Middle Proxy?

Mitm=man in the middle

Reply on news.ycombinator.com