<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:dc="http://purl.org/dc/elements/1.1/"
	xmlns:atom="http://www.w3.org/2005/Atom"
	xmlns:sy="http://purl.org/rss/1.0/modules/syndication/"
		>
<channel>
	<title>Comments on: Protecting your data: the broken drives edition</title>
	<atom:link href="http://blog.evernote.com/tech/2012/09/25/protecting-your-data-the-broken-drives-edition/feed/" rel="self" type="application/rss+xml" />
	<link>http://blog.evernote.com/tech/2012/09/25/protecting-your-data-the-broken-drives-edition/</link>
	<description>The Care and Feeding of Elephants</description>
	<lastBuildDate>Thu, 09 May 2013 03:19:03 +0000</lastBuildDate>
	<sy:updatePeriod>hourly</sy:updatePeriod>
	<sy:updateFrequency>1</sy:updateFrequency>
	<generator>http://wordpress.org/?v=3.5.1</generator>
	<item>
		<title>By: Chris L</title>
		<link>http://blog.evernote.com/tech/2012/09/25/protecting-your-data-the-broken-drives-edition/#comment-1679</link>
		<dc:creator>Chris L</dc:creator>
		<pubDate>Sun, 24 Mar 2013 03:53:29 +0000</pubDate>
		<guid isPermaLink="false">http://blog.evernote.com/tech/?p=392#comment-1679</guid>
		<description><![CDATA[You will require a password (or some key) when you load the encrypted filesystem. Obviously there is no need for the root filesystem to be encrypted.

And there&#039;s a trivial way to get the key at that point: put it on a cheap, tiny usb key, which is left permanently plugged into the server. At boot, read the password from the USB key, and then use it to mount the encrypted filesystem.

The point is not to protect the key from exposure should someone access the system. Indeed, at that point the data is already compromised, since the filesystem is mounted. The point is to protect the data after the disk is removed, which this does safely. Just don&#039;t dispose of the USB key AND the disk together. ;)

[Of course, we can get more complicated and avoid the USB key disposal problem by having a keyserver. One solution would be a REDIS instance, where the key for each machine can be supplied (once the new machine has its network interface up).]

The point is that it&#039;s not difficult to get this working, saves a lot of hassle and expense dealing with disk disposal, and is a lot more foolproof from a security perspective.]]></description>
		<content:encoded><![CDATA[<p>You will require a password (or some key) when you load the encrypted filesystem. Obviously there is no need for the root filesystem to be encrypted.</p>
<p>And there&#8217;s a trivial way to get the key at that point: put it on a cheap, tiny usb key, which is left permanently plugged into the server. At boot, read the password from the USB key, and then use it to mount the encrypted filesystem.</p>
<p>The point is not to protect the key from exposure should someone access the system. Indeed, at that point the data is already compromised, since the filesystem is mounted. The point is to protect the data after the disk is removed, which this does safely. Just don&#8217;t dispose of the USB key AND the disk together. <img src='http://blog.evernote.com/tech/wp-includes/images/smilies/icon_wink.gif' alt=';)' class='wp-smiley' /> </p>
<p>[Of course, we can get more complicated and avoid the USB key disposal problem by having a keyserver. One solution would be a REDIS instance, where the key for each machine can be supplied (once the new machine has its network interface up).]</p>
<p>The point is that it&#8217;s not difficult to get this working, saves a lot of hassle and expense dealing with disk disposal, and is a lot more foolproof from a security perspective.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Alexei Rodriguez</title>
		<link>http://blog.evernote.com/tech/2012/09/25/protecting-your-data-the-broken-drives-edition/#comment-816</link>
		<dc:creator>Alexei Rodriguez</dc:creator>
		<pubDate>Tue, 05 Feb 2013 21:55:56 +0000</pubDate>
		<guid isPermaLink="false">http://blog.evernote.com/tech/?p=392#comment-816</guid>
		<description><![CDATA[Chris, good question. It is my understanding of the Linux software based full disk encryption (FDE) that once put in place you are then prompted for the passphrase at boot time.  While this may be fine for individual use, it does not scale to hundreds / thousands of systems. It is a solvable problem, but non-trivial. 

Encryption at the raid controller layer is another option which is interesting but we have not thoroughly tested yet.]]></description>
		<content:encoded><![CDATA[<p>Chris, good question. It is my understanding of the Linux software based full disk encryption (FDE) that once put in place you are then prompted for the passphrase at boot time.  While this may be fine for individual use, it does not scale to hundreds / thousands of systems. It is a solvable problem, but non-trivial. </p>
<p>Encryption at the raid controller layer is another option which is interesting but we have not thoroughly tested yet.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Chris L</title>
		<link>http://blog.evernote.com/tech/2012/09/25/protecting-your-data-the-broken-drives-edition/#comment-757</link>
		<dc:creator>Chris L</dc:creator>
		<pubDate>Mon, 28 Jan 2013 04:42:44 +0000</pubDate>
		<guid isPermaLink="false">http://blog.evernote.com/tech/?p=392#comment-757</guid>
		<description><![CDATA[Why not simply encrypt the content on the drives? That way you don&#039;t need to destroy anything: the data on the disk is inaccessible when the drive is separated from the servers.

You would have to manage distribution of the encryption key(s), but that&#039;s not a big problem - there are many examples of doing this in a mostly secure way just a short google search away. And the security of the key doesn&#039;t have to be overly excessive. One only has to ensure it&#039;s not published outside of the network, or stored unencrypted onto a drive. (If someone hacks into the network to steal the key, then the data is already at risk anyway).

As for performance: all modern CPUs contain hardware support for encryption functions, and all modern OS&#039;s have support for encrypted filesystems that leverage the CPU. The net result is that you can trivially encrypt the content of disks, with minimal performance overhead.]]></description>
		<content:encoded><![CDATA[<p>Why not simply encrypt the content on the drives? That way you don&#8217;t need to destroy anything: the data on the disk is inaccessible when the drive is separated from the servers.</p>
<p>You would have to manage distribution of the encryption key(s), but that&#8217;s not a big problem &#8211; there are many examples of doing this in a mostly secure way just a short google search away. And the security of the key doesn&#8217;t have to be overly excessive. One only has to ensure it&#8217;s not published outside of the network, or stored unencrypted onto a drive. (If someone hacks into the network to steal the key, then the data is already at risk anyway).</p>
<p>As for performance: all modern CPUs contain hardware support for encryption functions, and all modern OS&#8217;s have support for encrypted filesystems that leverage the CPU. The net result is that you can trivially encrypt the content of disks, with minimal performance overhead.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: David B. Donaldson</title>
		<link>http://blog.evernote.com/tech/2012/09/25/protecting-your-data-the-broken-drives-edition/#comment-606</link>
		<dc:creator>David B. Donaldson</dc:creator>
		<pubDate>Fri, 12 Oct 2012 01:37:27 +0000</pubDate>
		<guid isPermaLink="false">http://blog.evernote.com/tech/?p=392#comment-606</guid>
		<description><![CDATA[Hey, I&#039;m taking this SSD thing seriously.

SSD drives are very easy to destroy, and there are many decent ways to do it.

Method 1: Pull the cover, ground the SSD&#039;s negative power lead and apply multiple ultra-high-voltage sparks to the NVRAM chips using a high-voltage spark generator.

There are lots of commercially available spark generators which will work perfectly for this. Some even run on 120 volts, are trigger operated and include full isolation so no gloves or external power supply are needed.

I can hear the buzzzzz right now!

Allow the spark to go through the center of each NVRAM chip&#039;s case. This will totally destroy the chip.

There will be NO accessible information left on the SSD.

Note: Ventilate the area with a low velocity fan to disapate any O-zone vapors.

Total time required, in a production environment, well under 5 minutes per drive.

Method 2: Purchase a small electric ceramics oven (kiln). Place multiple SSD&#039;s as tightly packed as you like, into the kiln and heat it into a 1,200 Celsius 30 minutes cycle. All cooked. No more data - ever.

Silicon junctions in integrated circuits begin to fail above 150 degrees Celsius.

Method 3: Open the SSD and solder #22, 600 volt rated stranded leads directly to the + and Ground rails feeding the NVRAM chips. (This bypasses the over-voltage protection of the 5-volt power supply)

In an electrically isolated environment, connect the Ground lead to AC Neutral, and the + lead to the AC hot lead that is in series with a conventional 120 volt, 100 watt incandescent light bulb. This circuit should float (neither side connected to earth ground) and the power should be fed through a double-pole switch fed from a GFI outlet or circuit with a GFI circuit breaker installed.

Place the device to be burned out under or in a protective cover (i.e. a plexiglass box) on a concrete floor. This prevents any scintillation from hitting the operator.

Apply the power for 30 seconds. Allow to cool for 60 seconds, and there will be no more data.

A very small AC welder can also be used.

Method 4: Buy an old 20-ton stamp press, place an SSD drive between the jaws with the lower jaw flat, and the upper jaw fitted with a carbide tipped pin-grid array, and wearing the appropriate eye-wear depress the activator.

Bah-Dah-Boom, no more data.

This simple approach is fool proof, as it will penetrate and crack the case of each NVRAM chip, breaking the ceramic wafer inside.

------------------

Any of these methods can be automated. Methods 1, and 3 can be performed in seconds, quite possibly without having to open the SSD case.

Method 4 also is very, very fast, and does not require case removal or special handling. You will have to sweep the floor afterwards however.

Write me if you&#039;d like a tested prototype developed for you! I can even have it certified!]]></description>
		<content:encoded><![CDATA[<p>Hey, I&#8217;m taking this SSD thing seriously.</p>
<p>SSD drives are very easy to destroy, and there are many decent ways to do it.</p>
<p>Method 1: Pull the cover, ground the SSD&#8217;s negative power lead and apply multiple ultra-high-voltage sparks to the NVRAM chips using a high-voltage spark generator.</p>
<p>There are lots of commercially available spark generators which will work perfectly for this. Some even run on 120 volts, are trigger operated and include full isolation so no gloves or external power supply are needed.</p>
<p>I can hear the buzzzzz right now!</p>
<p>Allow the spark to go through the center of each NVRAM chip&#8217;s case. This will totally destroy the chip.</p>
<p>There will be NO accessible information left on the SSD.</p>
<p>Note: Ventilate the area with a low velocity fan to disapate any O-zone vapors.</p>
<p>Total time required, in a production environment, well under 5 minutes per drive.</p>
<p>Method 2: Purchase a small electric ceramics oven (kiln). Place multiple SSD&#8217;s as tightly packed as you like, into the kiln and heat it into a 1,200 Celsius 30 minutes cycle. All cooked. No more data &#8211; ever.</p>
<p>Silicon junctions in integrated circuits begin to fail above 150 degrees Celsius.</p>
<p>Method 3: Open the SSD and solder #22, 600 volt rated stranded leads directly to the + and Ground rails feeding the NVRAM chips. (This bypasses the over-voltage protection of the 5-volt power supply)</p>
<p>In an electrically isolated environment, connect the Ground lead to AC Neutral, and the + lead to the AC hot lead that is in series with a conventional 120 volt, 100 watt incandescent light bulb. This circuit should float (neither side connected to earth ground) and the power should be fed through a double-pole switch fed from a GFI outlet or circuit with a GFI circuit breaker installed.</p>
<p>Place the device to be burned out under or in a protective cover (i.e. a plexiglass box) on a concrete floor. This prevents any scintillation from hitting the operator.</p>
<p>Apply the power for 30 seconds. Allow to cool for 60 seconds, and there will be no more data.</p>
<p>A very small AC welder can also be used.</p>
<p>Method 4: Buy an old 20-ton stamp press, place an SSD drive between the jaws with the lower jaw flat, and the upper jaw fitted with a carbide tipped pin-grid array, and wearing the appropriate eye-wear depress the activator.</p>
<p>Bah-Dah-Boom, no more data.</p>
<p>This simple approach is fool proof, as it will penetrate and crack the case of each NVRAM chip, breaking the ceramic wafer inside.</p>
<p>&#8212;&#8212;&#8212;&#8212;&#8212;&#8212;</p>
<p>Any of these methods can be automated. Methods 1, and 3 can be performed in seconds, quite possibly without having to open the SSD case.</p>
<p>Method 4 also is very, very fast, and does not require case removal or special handling. You will have to sweep the floor afterwards however.</p>
<p>Write me if you&#8217;d like a tested prototype developed for you! I can even have it certified!</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Alexei Rodriguez</title>
		<link>http://blog.evernote.com/tech/2012/09/25/protecting-your-data-the-broken-drives-edition/#comment-605</link>
		<dc:creator>Alexei Rodriguez</dc:creator>
		<pubDate>Mon, 01 Oct 2012 01:23:41 +0000</pubDate>
		<guid isPermaLink="false">http://blog.evernote.com/tech/?p=392#comment-605</guid>
		<description><![CDATA[Although I have not seen the &quot;not for children&quot; safety advisory on these devices, I am pretty sure they are not child safe :)]]></description>
		<content:encoded><![CDATA[<p>Although I have not seen the &#8220;not for children&#8221; safety advisory on these devices, I am pretty sure they are not child safe <img src='http://blog.evernote.com/tech/wp-includes/images/smilies/icon_smile.gif' alt=':)' class='wp-smiley' /> </p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Thalia</title>
		<link>http://blog.evernote.com/tech/2012/09/25/protecting-your-data-the-broken-drives-edition/#comment-604</link>
		<dc:creator>Thalia</dc:creator>
		<pubDate>Fri, 28 Sep 2012 11:57:20 +0000</pubDate>
		<guid isPermaLink="false">http://blog.evernote.com/tech/?p=392#comment-604</guid>
		<description><![CDATA[Give them to my sister&#039;s two-year old and look away for five seconds. 

That ought to do it.]]></description>
		<content:encoded><![CDATA[<p>Give them to my sister&#8217;s two-year old and look away for five seconds. </p>
<p>That ought to do it.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Alexei Rodriguez</title>
		<link>http://blog.evernote.com/tech/2012/09/25/protecting-your-data-the-broken-drives-edition/#comment-603</link>
		<dc:creator>Alexei Rodriguez</dc:creator>
		<pubDate>Tue, 25 Sep 2012 19:01:07 +0000</pubDate>
		<guid isPermaLink="false">http://blog.evernote.com/tech/?p=392#comment-603</guid>
		<description><![CDATA[The guidance on SSD disposal is still evolving, but you are correct that published &lt;a href=&quot;http://static.usenix.org/events/fast11/tech/full_papers/Wei.pdf&quot; rel=&quot;nofollow&quot;&gt;research&lt;/a&gt; suggests that degaussing is not 100% effective. Thankfully the nature of failed SSDs is such that the device itself is still able to be accessed and the secure wipe instructions applied. As for the destruction: the machine does bend discs with platters, but it does have a bit more of a destructive effect on the SSD, as there is not the same resistance. We are doing our own testing of the different failure modes / wiping methods in order to get to a fool proof method. Some have suggested using a microwave oven (remains to be tested). Ideally we could have a modified wood chipper, but the data center did not approve it. We have also looked at having on-site shredder visits, but have not found any suitable local providers; they all want to take the drives and shred them at their offices.]]></description>
		<content:encoded><![CDATA[<p>The guidance on SSD disposal is still evolving, but you are correct that published <a href="http://static.usenix.org/events/fast11/tech/full_papers/Wei.pdf" rel="nofollow">research</a> suggests that degaussing is not 100% effective. Thankfully the nature of failed SSDs is such that the device itself is still able to be accessed and the secure wipe instructions applied. As for the destruction: the machine does bend discs with platters, but it does have a bit more of a destructive effect on the SSD, as there is not the same resistance. We are doing our own testing of the different failure modes / wiping methods in order to get to a fool proof method. Some have suggested using a microwave oven (remains to be tested). Ideally we could have a modified wood chipper, but the data center did not approve it. We have also looked at having on-site shredder visits, but have not found any suitable local providers; they all want to take the drives and shred them at their offices.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Jorge Vasquez</title>
		<link>http://blog.evernote.com/tech/2012/09/25/protecting-your-data-the-broken-drives-edition/#comment-602</link>
		<dc:creator>Jorge Vasquez</dc:creator>
		<pubDate>Tue, 25 Sep 2012 06:42:18 +0000</pubDate>
		<guid isPermaLink="false">http://blog.evernote.com/tech/?p=392#comment-602</guid>
		<description><![CDATA[How do you destroy SSDs? Degaussing and physically bending them wouldn&#039;t be effective in erasing data inside the flash chips.]]></description>
		<content:encoded><![CDATA[<p>How do you destroy SSDs? Degaussing and physically bending them wouldn&#8217;t be effective in erasing data inside the flash chips.</p>
]]></content:encoded>
	</item>
</channel>
</rss>
