Adventures in S/MIME – Getting a certificate

As a big fan of crypto, it has always rather amazed me that S/MIME hasn’t had more success. We hear many accounts of users cannot handle the certificate management required to make something like PGP or S/MIME work. I have been doing some experiments and have concluded that we got it wrong; we are blaming the user when the right person to blame is ourselves. This post describes the first step of success, what it takes to “get a key”.

S/MIME Encryption key – purchase

The end-goal is S/MIME email. This requires getting an encryption certificate, issued by a provider that has a common root of trust. Commercial certificate authorities (CA) are the answer here as their statements of authenticity will root up to validate trust by the vast majority of the worlds computers.  There are many providers to choose from, I chose Entrust. For $20 / year, they will issue me a personal certificate that I can use for S/MIME. How do I know it will work for S/MIME? Simple, it says so, right there in the buy a cert page

Entrust – compliments

Entrust will receive a few rocks in this post, but I will also follow with compliments on their support group and their willingness to help me resolve what was an original fail. For $20 – support costs, they lost money on my sale, but they stood by it and helped me succeed. For that, I write this post so that others will also succeed; hopefully this returns the favor to the kind folks on support line that help me get to success.

Adventures in buying a certificate

For Entrust, the site to start is Visit there, pay your money and out pops a certificate that you can use for S/MIME email. Or so I thought. I am using Windows 10 and I think the site is making assumptions to older versions of Windows and older web browsers.

Mistake #1 – I used Firefox. You must use Internet Explorer. The website doesn’t check this and if you do use Firefox, you will make it through the “pay money” part of the process, but will fail when it comes to the download the purchased certificate part. Turns out that the certificate sale at the end triggers an ActiveX activity to install the certificate into the Windows certificate store and from Firefox, this downloads as the public key part of the certificate only, so you end up with a certificate installed that will not allow you to do anything useful because you don’t have the private key.

Solution A: Contacted Entrust support line, and they issued me a discount code to start over.
Why IE? The website FAQ implies that public key private key pair are generated at Entrust and then they send you the certificate file encrypted with the password you supply during purchase. This isn’t how it works. It is how the website FAQ says it works, but isn’t how it really works. How it works in reality is better, but requires running code client side to do the key generation.

I do not want private key to have ever been in the possession of CA, or anyone except me and neither do they. We want to send them the public key, and have them sign it. This means that the key pair have to be generated on our PC and the public part sent up for signature and then back down to be reassembled and stored as our private key containing certificate which gets written into our system.

To accomplish that, they need ActiveX in the browser or some means to run code client side and with that, they depend on Internet Explorer ability to run ActiveX extensions.

Mistake #2 – ActiveX is rightly blocked by Internet Explorer

I’m on Windows 10. IE security gets better all the time and one thing it has done to get better is turning off the ability for any old website to run ActiveX controls (native code client side). Version of Internet Explorer that ships with Windows 10 does not permit ActiveX for non-trusted sites.

Solution: Temporarily add to the list of trusted sites in Internet Explorer.

Refresh webpage and think you’ll be good to go. No. Still fails.

Mistake #3 – ActiveX control from Entrust is not properly signed (or not signed at all)
Even with the sales URL added to trusted sites in IE, the browser still refuses to run the ActiveX control because it is not signed. That part by the way is kinda interesting for code from a CA, but save for another day.

Solution: Already willing to lower shields in talking to this website, another step is required to lower shields again to run non-signed ActiveX. This is global for all sites in Trusted Sites, so must be done only temporarily.

Refresh page and the ActiveX control will run and you can complete the purchase.
It will start with a warning about running the ActiveX control

You bet, proceed!

Warning from Windows – an application is wanting to create a private key! Yippie!

Security level at medium means that you will get prompted to approve access to the private key, but you will not have to enter an access password on each usage. Medium is great, next.

Then you are prompted with a confusing question on what you are for the certificate. You are the authorization contact – the one of highest authority over the cert, king of your own world.

And finally a happy “you’re done” screen — DO NOT CLOSE THE BROWSER!

Do not close the browser

You think you’re done, you’re not. The private key is still in memory inside the ActiveX control inside the browser. Leave it open. Go check email.

Email arrives – has a link to proceed to the next step

Do not click the link – that would just put you back in Firefox or Chome or your other browser that probably isn’t the same browser instance that is running the ActiveX control. Instead, Right mouse button, COPY the link into clipboard.

Go back to the Internet Explorer browser that did the purchase. In theory a new “tab” in that browser would be okay. I did not do that, I pasted into the URL window for the same tab that indicated I had completed purchase. This one I am sure can still see the private key. Paste the URL and press enter to browse to the site.

License agreement

Here, you enter the password that you provided earlier – so that the certificate host side can be downloaded and decrypted.

Happiness – Certificate installation happening, being commanded from the Entrust ActiveX control.

And done. The certificate is installed! Purchase process is complete.

Raise your shields!

The trusted sites addition done during purchasing should be removed and the default security level for sites on the trusted sites should be returned.

Alt-Tools, Options, Security: Remove site from trusted sites and raise the security bar for trusted sites back to the default, Medium is where it was before I had to dial it down.

Backup to off computer storage

The certificate is installed onto the computer, but at this moment, that is the ONLY copy. If the computer were to die, access to that certificate will be forever lost. Must back up as a standard part of purchase.

Still in Internet Explorer

Alt-Tools, Options, Security, Content, Certificates

Sends you into the Certificate Export Wizard.
Go through the first screen and on the export screen, tell it YES, include the private key.

It requires you to add a password, think one up, write it down, export the file. Store on a USB drive, put it with your backups…

And you’re done! Right?


And you’re done! Right? Answer: You’re not done.

You think you’re done, but you are not. You have successfully purchased and installed a certificate that theoretically can be used for S/MIME, but in reality all you have is a personal certificate that can be used for web browsing. You can now use public/private key crypto via this certificate to mathematically prove your identity to a website. Thankfully this will not happen without your approval, so you can leave this certificate in the Internet Explorer key store and go back to using Firefox who has no vision to it.

As for me, I bought this S/MIME certificate with the goal of sending and receiving encrypted email and so far, it won’t do that. Part 2 of this series will describe how to get Microsoft Outlook to utilize this certificate for S/MIME, link.

Joe Nord

(Originally published Jan 22, 2016)

Android apps on Windows Phone is like Windows on OS/2

Steve Ballmer is paraphrased in this ZDNet article saying “the company needs to ensure Windows Phone handsets can run Android apps”.  For a guy who spent more than 5 years writing system code for OS/2, the parallels to WinOS2 are pretty interesting.   Here’s the lesson: The operating system must stand on its own, or your just postpone it’s death.

“A better DOS than DOS and a better Windows than Windows”! 

With the 80836 supporting versions of OS/2 starting with 2.0 in 1992, the company did an absolutely excellent job running DOS applications in MVDM virtual machines.  This provided the required legacy support for DOS applications while the native side of the OS provided developers a platform to build rich applications that could fully exercise the systems 32-bit world.  OS/2 had an impressive multitasking kernel and a great TCP stack and could have been a great platform for application developers.  The applications never came; why?  A lot of reasons actually, but a big one was because IBM made the fantastic blunder to also build support for Windows 3.1 applications.

Once IBM built “adequate” system support for running Windows 3.1 applications on OS/2, ISVs now had zero motivation to write native applications and the native applications were not implemented, or were implemented poorly with customers instead opting to use the Windows applications on the OS/2 system and at the end, that operating environment just didn’t make sense anymore.

Development team gets distracted

Meanwhile, the operating system development team were spending tremendous effort to make Windows applications run inside a virtual machine.  Yes, Windows 3.1 was just a 386 DOS extender application running with a DOS boot loader, and given a MVDM supporting system already, running this big app wasn”t impossible.  It still took work though, real work!.   Work that ultimately included dragging me away from OS/2 native system work and into a world of writing what we today call paravirtualized drivers to send Windows 3.1 audio operations across to the native OS/2 multimedia system for processing.  Time that I SHOULD have been building the greatest audio and video processing system in the world, or perhaps just getting more device support for the native system.

Bottom line, it’s 20 years later.  Microsoft has become IBM and Google is playing the part of Microsoft.  If this “run Android on Windows” strategy actually proceeds, Android developers will have no motivation to write native applications for Windows phone handsets, and the operating system will ultimately die.

Everything that is old, is new again,

Joe Nord

Originally published Dec 3, 2015 11:12am

Spelling “lave” backwards is “eval”

Originally published Dec 24 2014

Found an unusual PHP file hiding in the root directory of my personal website, “s-g.txt”.  The file contains PHP code and last line contains “lave” which for some reason, the human brain quickly converts into “eval” and that’s suspicious, so it’s time to tear this apart. 

Note: Despite my attempts to post the original backward script here, the blog system rejects, identifying it as incoming evil, which it is.   It is below without scrambling.

First thoughts

The function names start with “wp”.  Once upon a time, I had hosted my personal blog via WordPress on my own website, but I got spam comments constantly so deleted the entire blog and removed WordPress.  Could this be leftovers from the removal of that blog?  When was the file created?

 "ls -l s-g.txt"
-rwx---r-x 1 joenord inetuser 807 Aug 13 02:52 s-g.txt

Okay, on August 13, 2014 at 02:52am Phoenix time, someone or some program did evil to my site.  A bit surprising though is the file permissions, readable to the world makes sense, but executable to the world, that’s odd.  I later learn that this is happening by default for all txt files created, so perhaps not related.

Getting back to execute permissions, the contents are PHP which means that execute permissions should not be needed for the contents to be read by the PHP engine so whether execute permissions are there or not, it will run if handed to PHP.

Next item, what does that “create_function” do?

Create_function – Create an anonymous (lambda-style) function

string create_function ( string $args , string $code )

This function internally performs an eval() and as such has the same security issues as eval().

Yeah, that’s bad.  More information on create_function, Parameters

Usually these parameters will be passed as single quote delimited strings. The reason for using single quoted strings, is to protect the variable names from parsing, otherwise, if you use double quotes there will be a need to escape the variable names, e.g. \$avar.

 Survey of the suspicious file says, there are double-quotes, more not good. Time to inspect the actual program operation.  Two lines of PHP code and I think I’m looking at an obfuscated ‘C’ contest.  The contents of the file are scrambled by reversing all characters in the function and storing the characters base64 encoded.  Borrowing some help from PHP, strrev does the first step of reversal, producing…


Close, but what I really want is the decoded form of the gibberish.  Base64_decode() and a bit of hand editing produces

$wp_function_initialize = create_function('$a', eval("{
     if(preg_match('/propecia|finasteride/i', $uniq_ref) > 0)
    }  elseif(preg_match('/cialis|cialas|cilis|tadalafil|cialis|ciallis|cialiss|cials/i', $uniq_ref) > 0)
        header("Location:");    // this line actually referenced blue pill word, but blogging system won't let it get posted

 Yes, evil

Yup, this could explain all those advertisements that were showing up in the comments of my blogs.   Keep inspecting.

What does that header() call do?  Documentation says that the location keyword is used to transfer web browser to another website and then return control to the calling PHP code, where a call to exit exists to end the script.

If operates as documentation says, then this will permit evil doer to adjust the spam content well after infecting the website.  Possibly though, if the PHP interpreter reaches out, gets raw html and then processes itself, then here’s a chance where arbitrary execution could occur.  Either way, it’s not good and has to go.  

.txt files – These are not supposed to execute as PHP

The website is hosted at Godaddy using their Linux hosting, txt file extension fetched from a web browser would normally not be processed as PHP.  Try it and Survey Says, that they are not being processed as PHP – so there is no security issue, presently.  ??  Paranoia and diagnosis for nothing?  Something tells me there is more to this story and if you know, please add in the comments.


Armed with information from tearing the script apart, I find the issue detailed here, where it is shown as the WordPress issues that GoDaddy experienced earlier in this year (2014).  Assume for the moment that GoDaddy neutered the script – why the this garbage still in the root directory of my website!

Actions to clean, remove suspect lines from .htaccess and erase suspect files.  “Eli” on this blog has a tool for securing WordPress.  If ever install WordPress again, should consider installing Eli’s defense plugin

Cleaning up

Looks like GoDaddy already removed the evil, but left some files behind.

.htaccess has extra items, these should be erased or commented

# RewriteEngine On
# RewriteCond %{HTTP_USER_AGENT} (google|yahoo|msn|aol|bing) [OR]
# RewriteCond %{HTTP_REFERER} (google|yahoo|msn|aol|bing)
# RewriteRule ^([^/]*)/$ /starting.php?p=$1 [L]
Extra files exist in the html root, these should be erased

Bottom line Today (Dec 24 2014), I do not have an issue – at least I do not have this issue.  On August 13th 2014, I did though and now, 4 months later, I spend a few hours dismantling the scripts to see what’s happening.  It was entertaining, so I share for all

Why Tiger Woods was penalized 2 strokes at the 2013 Masters

(Originally published Oct 24 2013)

At the 2013 Masters Golf Tournament, Tiger Woods nearly holed an approach shot on the 15th hole – only to have it bounce off the flag and go back into the water. Tough break!  A number of friends have asked me to explain the 3 strokes in penalty so I write it here for a wider audience.  Below is a graphic of the Augusta 15th green along with a markers for the where the shot was taken, and available drop locations.  The key point is that after the ball hit the flag, it did not bounce directly back toward the hitter, it went at an angle to the left and this changes the available drop locations.

Tiger Woods at Augusta #15, Masters 2013.

The USGA publishes the rules of the game and for today, we’re interested in rule 26-1, link, which discusses the players options for a drop after hitting a ball into a water hazard.  According to the rules, Tiger can drop in 3 places identified by black circles.  He cannot drop in the red-circle, but this is where he did drop and that is the source of all the chaos.

Legal drop spots – shown by number in the graphic

  1. Same spot as where hit the last shot.  Technically “ball as nearly as possible”
  2. Behind the water hazard, on a line (more precisely, a ray) originating at the flag and extending through the point where the ball LAST entered the hazard. This means it has to be on the same side of the canal as where the prior shot was taken and in this case, that location has a tree or is out of bounds, so it is not a candidate
  3. Drop area if designated and in this case, a drop area was defined so it was an option.  Precisely where that drop location lies on the #15 hole I do not know, but for today’s purposes, I’ve drawn it off to the right.  Key point, this option wasn’t selected, so it isn’t part of the present question

Note that for “2″, the line segment extends forever and the golfer may drop anywhere on that line so long as they are still on the course.  Reviewing options, the drop area was “wet”, so undesirable.  The original spot was not liked because it was a couple yards too close to be in Tiger’s sweet spot, so he selected a spot a few yards further away.

How Tiger got bit! 

NORMALLY when a golfer chips into the water right in front of the green, it “costs 2″ (The shot into the water +1 penalty) and you hit another from the same spot.  Or, you can back up as far as you want with a line on the flag. This however is not the rule, it is an abbreviation of the actual rule.  The abbreviation though “works” most of the time because the place the ball enters the water is usually directly between the original spot and the flag.  This means you can usually drop where you hit from, or anywhere further away, but this is only usually.

The rules don’t say where the ball entered the hazard, they say where the ball LAST entered the hazard and here, Tiger missed! In his case, the ball last entered the hazard AFTER it hit the flag stick.

Rules to your advantage

This going further back thing is not so well known.  It can though be very handy; for example trees are in the way.   Golf permits you to drop anywhere on the permitted line/ray and you can go as far back as you want, even past the tee-box!  With this, you can take the trees out of play.  Yes, you end up hitting a further shot, but it can be an easier shot.  This is 100% legal and knowing the rules can save you shots on your game.  This line/ray thing also makes it easy to back up a couple yards when you don’t like the lie where you hit a first shot into the water, usually.

That abbreviation though is NOT the rule of the game

The rules say to draw a line from the flag through the spot the ball LAST entered the hazard.  In tigers case, his ball crossed into the hazard TWICE.  First on the way to the green and second after it bounced off the flag.  In drawing the line/ray, Tiger mistakenly used the first passage into the hazard rather than the second and selected a position further away to get an advantage on his second try into the green.  This advantage is not permitted for 26-1 “a”, so it’s a 2-stroke penalty.

Interestingly, had the ball bounced off the flag directly back at Tiger this would work out to be a “same”.  The line/ray would go through the point of the original shot, but the ball did not bounce straight back!

Where things get complicated

Tiger completed the round, signed his card and turned it in.  Accounts say that the tournament committee were aware of the discrepancy before the round completed and decided that his couple yard movement was not an issue because they were both “same spot” as original.  That is, it provided no “advantage”, so the shot was the same.

BUT, in an interview after the tournament, Tiger noted that he purposefully DID select a spot further because he wanted the shot to be longer, to get it into his sweet spot for the second try, and it worked!    This triggered the rules committee, that he really wasn’t dropping in “same spot”, he was moving back on the line, but it wasn’t the right line!  Oh crap, what now!

The rules of the game say that if you turn in a score card with a score better than actual, you’re disqualified.  There is an exception added this year which was used in this case.  Good use in my view.

Twitter and news are abuzz with claims of preferential treatment.  Was there?  Hard to say, but here’s a counter statement that most golfers, most professional golfers, would have never been under the same microscope and their infraction would have gone unnoticed, even to themselves.

Also, the scorers following the group SHOULD have told Tiger of the 2 stroke infraction before he signed his card.  Technically, the golfer must ask for scoring assistance, so it falls back on the golfer, but really, would this same level of scrutiny existed for anyone other than the most watched name in the tournament?

Statute of limitations

The rules say nothing about a time limit of when a score card is “accepted” by the tournament.  At what point can a tournament no longer come back to say “you messed up”.  At some point, the score has to be finalized, or golfers will never be able to sleep at night as they recount every shot and every potential mistake of the day.

Right conclusion

The Masters and the rules of golf got it right, they charged Tiger Woods the 2 strokes for the violation, but he was not disqualified and was able to stay in the tournament.  His score was adjusted to be what it would have been if he had scored it correctly.  Above said, I am sure that Tiger Woods will not make this mistake again! 

Calls for Self-DQ

It is permissible for a player to turn in a score card with a score higher than what the tournament believes.  If they do, this score sticks.  With this, players can, and in the past they purposefully have self-imposed penalties that the tournament officials did not see.  This is consistent with the spirit of the game and it is a long tradition.  Was it appropriate here for Tiger to Self-DQ?  Debatable.  The penalty was imposed and Tiger was unaware of the proper score for the hole until it was imposed.  There was no chance to self-assign the penalty, only a chance to self-drop from the tournament, which really wasn’t warranted.

What should have happened

In professional tournaments, rules officials follow the players around the course and are available for in-round questions.   A small inquiry to the officials before the drop would have clearly defined Tiger’s options and would have saved him two-strokes and I’m sure, much anguish.

Given he didn’t ask, the scoring officials should have told him of his mistake and both could have assigned the penalty.  They didn’t tell him at the completion of the hole and they also did not bring it up at the end of the round before he could sign his card.

There’s blame to share, and the right outcome prevailed.

OS/2 Interrupt Handling

Written December 1, 2008. I recently received an inquiry regarding how OS/2 interrupts are handled and what is the correct action of a device driver upon being called by the OS/2 kernel. My first response was, you have got to be kidding me, the operating system has been dead for 10 years. The second response was to tell them the answer and now I write this blog so other folks might find it useful.
The failing scenario was a UNIX based sound API ported to OS/2 and the fact that it was dependent on “time” increasing during interrupt processing. The viewed behavior was that time would not increment so long as the device driver was doing work and this caused the sound library to come unglued. The solution was easy – TIME should increase while the device driver is processing an interrupt. You as the device driver writer should not prevent other device drivers from doing their work, especially an important device driver like the one inside the OS/2 kernel that keeps track of time.
The foundation of the problem was that the sound device driver in question was running its interrupt handler and that interrupt handler was preventing the dispatch of other interrupts. The solution: In the device driver interrupt handler, you should VERY EARLY enable further interrupts. This sounds like something you shouldn’t do, but you should. Example code describes better than words.

// Called by OS/2 kernel (interrupt dispatcher)
// On entry:
// DS is already set
// Interrupts are disabled
// On exit:
// We do not have to preserve the general purpose registers.
// We must clear the carry flag to tell the kernel that it
// was our IRQ.

void _interrupt IRQHandler (void)
   BYTE irqFlag;

   // Determine why the device generated the IRQ
   irqFlag = codecRead (...);

   if (it wasn't us)
       // Set carry flag to tell the OS/2 kernel that it 
       // wasn't ours and return to the kernel (iret).
       // Side note: An interrupt that is dispatched, but that
       // has zero device drivers claim responsibility will be
       // masked off by the OS/2 kernel before interrupt
       // processing is completed.
       Code omitted;

   // Acknowledge the device interrupt
   // In a level triggered world (PCI), the device stops pulling
   // on the interrupt line. Other devices can still be pulling.
   codecWrite (...);

   // Enable higher priority interrupts
   // Omitting this was the bug in the inquiry I received.
   // Enabling interrupts at the CPU does not mean that you will
   // be reentered. Quite the opposite, you WON'T be reentered
   // until you tell the 8259 interrupt controller that you have
   // completed processing this interrupt level.
   // By enabling interrupts at the CPU, what you are doing is
   // enabling the dispatch of "higher priority" interrupts where
   // priority is determined by the PIC.
   // In this example case, IRQ-8 (Timer) is higher priority
   // than IRQ-A (PCI). 
   // With the addition of the enable interrupts at the CPU, 
   // the timer was able to fire and "time" advances.

   // Do heavy lifting of moving data and otherwise doing the
   // work of pulling data from the device.
   // Depending on the architecture of the device, it may be
   // necessary to pull/push the data before acking the interrupt
   // at the device.
   Code omitted;

   // Time to return to the kernel
   // Prevent nesting by disabling interrupts at the CPU.
   // Kernel dispatcher will reenable interrupts when we return
   // and it is possible that it will again immediately call us.
   // This is okay because the stack will unwind before next call

   // Prevent interrupts by blocking all of them at the CPU

   // Finally - ack the interrupt at the 8259 PIC.
   // The PIC will send interrupts to the CPU, but it won't see
   // them, yet.
   DevHelp_EOI (codec_int);

   // Tell the kernel that we handled this IRQ so that it will
   // skip calling any other device drivers that may be
   // registered for this interrupt.

   // Observe that interrupt flag is still clear, this is
   // critical to prevent reentry once the EOI was commanded.
   // iret will be generated by the compiler as part of the
   // function return statement 
   // (dictated by the _interrupt prefix).
   // Kernel will re-enable interrupts (sti) soon after 
   // our return.