According to researchers from Outpost24, there are some flaws in the TCP/IP protocol that connects all systems on the internet. The flaws are so fundamental that every single system they tested could be brought down in a matter of seconds from a normal broadband connection. While these kinds of flaws are not exactly new, this problem is a very big deal. They were not able to find a fix on their own, and have contacted vendors to see whether they can find a solution:

Robert and Jack are waiting with no specific timeline to hear back from the affected TCP stack vendors. Think firewalls, OSes, Web-enabled devices, and so on. Yup, they’ll all need to be hardened, if the vendors can come up with a good solution to the problem¬†

DarkReading has a post with (some) details. but the article is (understandably) light on details. Techtarget has more info, and mentions that the flaw has been discovered years ago:

The problems, which were identified as far back as 2005, are not simply vulnerabilities in products from one or two vendors, but are issues with the ways in which routers, PCs and other machines handle TCP connection requests from unknown, remote machines. 

From the information I’ve been able to gather so far, the flaw is in the handling of new connections; apparently some systems keep sending packets after a single connection attempt, until the affected machine is rebooted. Another publication mentions that 10 packets per second can bring down an entire machine, so I think that this is an issue with SYN/ACK packets. New information is to be presented at the T2 event.

Update: there is a slashdot-article about this as well, and Robert E. Lee provides some corrections and additional information in this blog post.

Update 2: I’ve just posted a summary of all new information that has become available.