3 ms·
The slidedecks from two presentations are far more interesting and enlightening than the original post, which instead refers to a new batch of available data.
by asymptotic 16y ago
The slidedecks from two presentations are far more interesting and enlightening than the original post, which instead refers to a new batch of available data.
From DEFCON18, titled "An Observatory for the SSLiverse":
https://www.eff.org/files/DefconSSLiverse.pdf https://www.eff.org/files/DefconSSLiverse.pdf
From the 27th Chaos Communication Congress, titled "Is the SSLiverse a Safe Place?"
https://www.eff.org/files/ccc2010.pdf https://www.eff.org/files/ccc2010.pdf
The methodology is to nmap all IPs for open port 443s from a tiny three-machine cluster, and if the port is open launch a simple Python client that initiates an SSL handshake but stops before sending its own key. They then dump absolutely everything the server tells them into a MySQL database.
The purpose isn't just to pressure CAs; for me an enlightening and disturbing fact is just how many trusted CA certs there are. They observed 1,482 CA certificates from _651 organisations_ trusted by Windows and Firefox! 651!
There are examples of CAs signing "192.168.1.2" and "localhost", two CA certificates signed with 508-bit RSA keys, etc. Quite interesting.