Jeffh.net

I am playing around with konboot after first hearing about it on the Hak5 podcast.  If you haven’t heard about it, it is a boot disk you can create on a floppy, CD or USB drive (see Irongeek’s site for USB instructions). The disk changes the windows or Linux kernel on the fly while booting to allow you to bypass the login password.  In windows you just use any valid user name and a blank or garbage password, you will then be logged in as that user and can access all their files.  So naturally you will probably want to try the local administrator user which will allow you to access all files on the computer.  In Linux you will use the user name kon-usr and no password.  This will give you full root access on the Linux machine.

I have found it works well if the system is not muti-boot.   On my systems which are all multi-boot it would only work on one system and only on one OS which happen to be EEEbuntu.  My other laptop gave me error about the cylinder number being too high like the old school days of LILO where the boot image had to be below cylinder 1024. I am assuming this is the same issue as the days of old.  Also a caveat don’t use Konboot to login to a domain account on a computer that is connected to the network.  This will disable the account on the domain and won’t allow you to login.  In most environments user credentials are cached in case the network goes down.  So air gap the computer before using a domain account.

Remediation steps are fairly simple.  Lock the bios with a password and only allow the system to boot from the hard drive.  This should already be in the check list of task to perform when deploying a new PC. Since this type of threat isn’t new kon-boot just make it a little simpler to access the PC than loading up a live linux distro like knopix.  If there is any sensitive information on the hard drive encryption should be used of course since if someone steals the computer or hard drive its game over.  With the breach notification laws in most states that is not a fun proposition.

The Linux Cisco VPN client seems to try the first avalable interface to make a connection. In my case this was eth0 which is my wired port which isn’t connected to anything. In order to get it to work I had to down the interface with ifconfig eth- down then it would connect fine with my wireless interface eth1. I haven’t seen any documentation to see if there is a way to manually bind it to a specific interface, and there is no MAN page installed. At least –help gives some usage, which is heads and shoulders over the tsmc tivoli command line I was using today what a royal PITA. Just encase anyone needs it the command to restore a previous version of a file is “tsmc restore -todate=yyyy-mm-dd -inactive /path/to/file” the documentation gave the format for the date as mm/dd/yyyy and the command would just come back and say -todate not recognized. Not wrong date format no that would have saved Jeff an hour of goggling.

I live in California where the Secretary of State Debra Bowen has allowed The University of California to test the E-Voting systems used in here in California.  What they found was that all of them are hackable. I find the preliminary findings very scary but not at all surprising.  The legislature of California are not Computer Security Experts but they should have consulted with and let Security Professionals verify that they were secure before Certifying the systems.  It is inexcusable to allow what is arguably the most important Computer System in the country to be so insecure.  These machines are what decide our election.  If you could modify the election results you could execute a bloodless coup and no one would be the wiser.   No one could prove that an election had been stolen if there is no paper trail.

I will give you one scenario.  All the electronic voting machines are stored in some medium security warehouse with a few guards and 1 or 2 cameras.  A disgruntled technician from company XYZ that makes $20,000 a year maintains voting machines is payed some large sum to hack the companies voting machines.  He is admitted entry into the facility do do “maintenance” He turns them all on and one by one inserts a USB thumb drive that installs a virus on each machine.  He also updates the BIOS and changes the checksum check that is used to verify that the operating system is certified, just like he has to do when he installs any patch.  This virus will change the vote on 41% of the votes in the favor of candidate FOO.  Thus virtually insuring victory.  This virus then removes all traces of itself and restores the correct checksum back into the BIOS at 7:55PM on election day.  With the current systems that have no paper trail and no one would find out unless someone starts looking at the technicians large bank account.

I am not saying to go back to all paper ballets, to me this is even more error prone and hackable than E-voting.  Just make up some ballet boxes and switch them in transit to the registrars office.  No to me the problem is fairly trivial technologically speaking.  The solution is 3 fold.  First and most simple a paper copy must be printed out and verified correct by the voter.  Thus ensuring a backup in case the electronic version is lost or tampered with.

The second part of my proposed solution is PKI (Public Key Infrastructure).  Each voter should be assigned a private key, a 3D bar code would work nicely.  This key should be assigned completely at random and only good for 1 election so that voting patterns couldn’t be data mined and somehow connect to an individual, there should also be no record kept of the voters private key only their public key should be kept.  This private key should be encrypted with a master public key so that there is no way a voters private key and identity could be obtained by coping it in transit unless the person also had the master private key.  The voter then would then scan their bar code and the voting machine would verify that it is a legitimate key by decrypting the key with the master private key.  The voter would then vote and the results would be signed with the users private key, and also be printed out.  If the results were tampered with the signature would not match and it would be obvious that the results were tampered with.  If two different votes were signed with the same private key you would also know that the user voted twice and to trow out all other votes signed with the same signature.   The  public keys and paper copies should be escrowed so that recounts could be done for some defined time.

The last and most important is both the physical security and openness of the system.  The inputs should be protected with 3DES encryption so that only certified devices could be plugged in.  No uncertified devices should be allowed on the system.  No one should be able to plug a key board in.  The case should be wielded shut and painted with a heat sensitive paint so any attempted tampering would be apparent.  The system should be completely open source from the OS to the voting software so that all code can be freely audited.  It someone did manage to hack the system the paper backup should throw up the red flag that the system was compromised.  The system should also be on an encrypted EEPROM so that even if you get into the case you can’t modify the EEPROM with out the key.  The key should be on a physically separate  dongle that is  locked up  separate from the voting machine and must be present when booting the machine and removed thereafter.  If any dongle is lost then all the dongles will have to be replaced and the EEPROM reprogrammed so it is very important that they don’t go missing.  The systems should never be plugged into an open network.  The system that collects the votes should also have a certificate assigned to it and stored on the voting machines so that the voting machines don’t disclose voting results to an unauthorized system.

All of this technology is common and in use today.  We need to reform the voting system so that we implement this as soon as possible.  If we don’t we may soon find our country not in our control anymore.  I don’t go into all the details about the system but I would love feedback, what are the weaknesses in the system I described.  If done correct E-Voting is both more efficient and more secure than paper based voting alone.

I am redoing the IDS system at work.  The current IDS uses snort and BASE.  Base just isn’t able to keep up under the current load.  We got a new quad core server with 4 gigs of ram that should do nicely for base.  I also decided to redo the sensors at the same time.  They were both running FreeBSD and I wanted to go to CentOS 5 because I am much more familiar with it than FreeBSD.  I installed CentOS 5 and did a test run with tcpdump.   It just can’t keep up the kernel is dropping packets even when loging just the raw packets to a file.  The traffic isn’t huge its about 3000 packets per second (around 32 Mbps).  Even after tunning some kernel parameters and some network parameters it was still dropping packets.

From what I have read the FreeBSD network stack is a lot better at this type of thing.  I am testing another FreeBSD box to see if it can keep up without dropping packets.  If it can keep up then I am going to use freeBSD for the sensors and stick with CentOS for the BASE front end.

UPDATE:  FreeBSD took it like a champ 0 dropped packets, so it looks like its FreeBSD for snort