Scientists have developed free software that can prevent the techniques used by the Hacienda spy program to avoid a server being taken over.
Today, a group of journalists has reported the existence of the "Hacienda" spy program. According to this report, five western intelligence agencies are using the Hacienda software to identify vulnerable servers across the world in order to control them and use them for their own purposes. Scientists at the Technische Universität München (TUM) have developed free software that can help prevent this kind of identification and thus the subsequent capture of systems.
Port scanners are programs that search the Internet for systems that exhibit potential vulnerabilities. According to the report published today by journalists at Heise Online, Hacienda is one such port scanning program. The report says that this program is being put into service by the "Five Eyes," a federation of the secret services of the USA, Canada, the UK, Australia and New Zealand. "The goal is to identify as many servers as possible in other countries that can be remotely controlled," explains Dr. Christian Grothoff, Emmy Noether research group leader at the TUM Chair for Network Architectures and Services.
Source available at: https://gnunet.org/knock
(Score: 2) by frojack on Tuesday August 19 2014, @11:21PM
In a typical port knock scenario, you send a packet to host:portA maybe followed by several other pairs host:portX host:portM ...
When your port knock setup sees these connection attempts on closed ports, it denies them as usual, but it remembers the sequence of their arrival, and the time in which they arrived. If they arrive in the proper sequence, and in the proper time your iptables will open/listen on some other port, say SSH(22) for X minutes/seconds.
So SSH only becomes available AFTER you a given sequence of other ports in a period of time.
This isn't that hard to outsmart, because you can't arbitrarily try to reject the sequence if some other connection attempt arrives between say host:portA and host:portX, because there are enough random connection attempts happening all the time that resetting the sequence for a random connect would amount to a self-denial of service.
Port Knocking has be fooled strictly by random knocks, but also by anyone in a position to watch packets being sent.
See: a couple explanations of why port knocking is not all that great Here [blogspot.com] and
Here [securitygeneration.com]
I din't mean to imply there is an "extra communication" channel. This new method is using part of a standard SYN packet, the first packet used when trying to connect, a part that has historically been ignored.
Cute animated SYN handshake demo here: http://www.inetdaemon.com/tutorials/internet/tcp/3-way_handshake.shtml [inetdaemon.com]
Coverage of the Sequence portion of a SYN packet here:
http://packetlife.net/blog/2010/jun/7/understanding-tcp-sequence-acknowledgment-numbers/ [packetlife.net]
Scroll down to "Sequence and Acknowledgment Numbers"
When a host initiates a TCP session, its initial sequence number in the first SYN packet is effectively random; it may be any value between 0 and 4,294,967,295, inclusive. It originally was designed to be zero in the standard, but because of historical bugs, nobody checked the first SYN, and all stacks allow it to be anything.
These guys are planning to make use of that, so that the first SYN packet can also carry some (not so secret) information
to pass to the Kernel.
You could combine this with standard port knocking, so not only do you have to get the host:sequence series of packets, but they all have to contain your special sequence numbers in a specific order.
But these guys never mention that, they are just using it to pass some additional bit of information that will tell the target to open up (listen on) the SSH port. That seems simplistic to me. Log some traffic, get some keys.
That said, they are offering some additional features that have never been used, such as being able to tell the target stack exactly how much of the subsequent data stream to "protect".
I haven't waded through all the code yet myself.
No, you are mistaken. I've always had this sig.
(Score: 1) by X1 on Wednesday August 20 2014, @07:35AM
Thanks for the thorough explanation, it is good context even if I understood all the basic theory of port knocking already.
Here is where you further enlighten me, though as I suspect, I still remain unimpressed by this new technique. I.e. this is clearly just adding complexity to the 'knock'. My BUGS BUNNY technique is just that taken to the comically illustrative extreme. Likewise, they are using the "historically ignored" bits as a new 'covert' 'extra communication channel'. Which could be portrayed to the non-security-expert (read: NPR listener) as some uber-terrorist-helping covert communication ability. However to people who understand the basics of Internet Protocol and how networks work, seems a lot less interesting. I.e. any ISP or network that wanted to prevent this kind of activity could easily just rewrite those bits as 0s. End of 'new terrorist communication threat'. I could still be wrong and maybe there is something landscape-changing here, but I don't grok it yet. I find it far more plausible that this is something the NSA has been using for themselves, flying underneath the ISP radar, or if need be, twisting ISP arms to have the channel ignored. Now however they can spin it to people who don't understand that it really isn't any new capability, as some sort of terrorist hacker threat to make people afraid of how Snowden educated all those terrorists. When in fact, it isn't fundamentally of any more use to terrorists than simple best-practices use of GPG/SSH/IPSEC/ETC.
Likewise, as to kaszz's description of Google inhouse network practices, it sounds like fundamentally no greater gain than properly managed IPSEC/SSH/router-packet-filtering-of-unexpected-traffic/etc.
Clearly I'm trying to derive the conclusion that there is "nothing to see here folks, move along and ignore the scary talk meant to incite fear in those who don't understand the issues". Which is sadly my conclusion of many 'security' articles on popular tech sites since the post-snowden propaganda wars have been raging...
(Score: 2) by frojack on Wednesday August 20 2014, @05:19PM
I pretty much agree with your assessment, and I'm not sure that kernel patch they offered is going to make it into any standard kernel, (and they only address linux, not MS/BSD/etc). It makes a bug in stack handling into a feature that people depend on, all without benefit of a RFC.
No, you are mistaken. I've always had this sig.