Sunday, 15 May 2016

The Chimp That Pimps And An Introduction to e.MMC Flash Memory Forensics

Pimpin Ain't Easy?

SANS is offering the top 3 referrers to its DFIR Summit 2016 website, an Amazon Echo smart speaker.
As of 11 May 2016, this Chimpy McPimpy was number 5 on the list.
Chimpy would very much like to win an Echo (echo, echo) so he can take it apart and share what forensic artifacts are left on the device.

The Echo is a smart speaker that can listen out for voice commands, play music, search the Internet and control Internet Of Shitty Things. Apparently, more than 3 million have been sold in the US since 2014.

Here's a (pretty meh) Superbowl commercial demonstrating some of the Echo's capabilities:

And here's the Wikipedia entry for the Amazon Echo just so monkey doesn't have to regurgitate any further (I already have enough body image issues).

The folks at Champlain College have also recently blogged about their Amazon Echo forensic research (here, here and here).
They have a report due out this month (May 2016).
From what this monkey can ascertain, their research focuses on network captures and the Amazon Echo Android App side of things. They also mentioned looking into "chipping off" the device but I'm not sure if this was a core part of their research as it wasn't mentioned in later posts.

So Monkey is proposing this - (if you haven't already) please follow this link to the SANS DFIR Summit website and if monkey manages to win an Amazon Echo, he will blog about getting to that sweet, sweet, echoey data from the internal Flash memory. See here  and here  for some background on Flash memory.

How do we know it uses Flash memory?
The awesome folks at have already performed a teardown which you can see here.

From's picture of the logic board (below), we notice the Flash memory component bearing the text SanDisk SDIN7DP2-4G (highlighted in yellow).

Amazon Echo's Logic board

Searching for the Flash storage component(s) on most devices (eg phones, tablets, GPS, answering machines, voice recorders) starts with Googling the various integrated circuit (IC) chip identifiers. The Flash memory component is normally located adjacent to the CPU (minimizes interference/timing issues).
In this case, the peeps have helpfully identified/provided a link to the 4 GB SanDisk Flash memory chip.
But if we didn't have that link, we would try Googling for "SanDisk SDIN7DP2-4G" and/or "SanDisk SDIN7DP2-4G +datasheet" to find out what type of IC it was.
According to this link - for the 4th quarter of 2015, Samsung's NAND revenue (33.8%) led Toshiba (18.6%), SanDisk (15.8%), Micron (13.9%), SK Hynix (10.1%) and Intel (8%). Other (smaller) manufacturers such as Phison, Sony, Spansion were not mentioned. Not sure how accurate these figures are but if you see one of these manufacturers logos/name on a chip, you have probably found a NAND memory chip of some kind (eg Flash, RAM).

Anyhoo, from the link that provided we can see the following text:
SDIN7DP2-4G,153FBGA 11.5X13 e.MMC 4.51
Here's what it all means:
153 FBGA (Fine pitched, Ball Grid Array) means there are 153 pin pads arranged in a standard way.
The 11.5X13 refers to the chips dimensions in millimetres.
The e.MMC 4.51 tells us the chip adheres to the Embedded Multi-Media Card (e.MMC) standard (version 4.51) for NAND Flash chip interfacing. We will discuss the e.MMC standard a little further on.

To double check's data link, we did some Googling and found this link which seems to confirm from multiple sites that the SanDisk Flash chip is 153 FBGA and 11.5 x 13.
Ideally, we would have found the actual datasheet from SanDisk but sometimes you just gotta make do ...

It is also worth noting that not all Flash memory chips are e.MMC compatible. Some devices may use their own proprietary NAND interface. Some chips might be NOR Flash (eg Boot ROM) and thus not really relevant to our quest for user data.
Additionally, the latest Flash memory chips may follow a newer (faster, duplex) standard called Universal Flash Storage (UFS). See here for more details on UFS.
So while it appears the days of e.MMC chips are numbered, there's still a LOT of e.MMC storage devices out there that can be potentially read.

When reading Flash storage for forensics, some key considerations are:
- Does it follow the e.MMC standard?
- Chip pin arrangement (number of pins and spacing)
- Chip dimensions (typically in mm)

The e.MMC standard is used by Flash memory chip manufacturers to provide a common infrastructure / command set for communicating. This way a board manufacturer can (hopefully) substitute one brand of eMMC chip with another brand (probably cheaper) of the same capacity. The standard focuses on the external eMMC chip interfacing and not the internal NAND implementation (which would be manufacturer specific). Having a e.MMC Flash chip makes reading a whole lot easier.

But don't just listen to me, JEDEC - the folks responsible for the eMMC standard (and UFS), state :
"Designed for a wide range of applications in consumer electronics, mobile phones, handheld computers, navigational systems and other industrial uses, e.MMC is an embedded non-volatile memory system, comprised of both flash memory and a flash memory controller, which simplifies the application interface design and frees the host processor from low-level flash memory management. This benefits product developers by simplifying the non-volatile memory interface design and qualification process – resulting in a reduction in time-to-market as well as facilitating support for future flash device offerings. Small BGA package sizes and low power consumption make e.MMC a viable, low-cost memory solution for mobile and other space-constrained products."

To get a copy of the e.MMC standard (free registration required), check out this link.

The e.MMC standard document provides this helpful diagram:

JEDEC e.MMC Electrical Standard v5.1

From this we can see that a "Device controller" handles any interfacing with the actual NAND storage ("Memory Array"). This includes things like reading/writing to NAND, paging, TRIM, error correction, password protection.

There are 4 signals/pins required when reading an e.MMC memory:
- CLK = Synchronizes the signals between the e.MMC chip and the "Host Controller" (ie CPU of device)
- CMD = For issuing commands/receiving command replies from/to the "Host Controller"
- DATA0 = For receiving the data at the "Host Controller"
- VCC / VCCQ = Power for the NAND memory / Power to the Device Controller. In some cases, this can be the same voltage (1.8 V)
- GND / VSS = Ground

It is not a co-incidence that these connections are also required for In-System Programming (ISP) Forensics. But that is probably a topic more suitable for a Part 2 (hint, hint).

We can see these pins labelled in this ForensicsWiki diagram of a BGA 153 e.MMC chip
BGA-153 Layout

Note: ForensicsWiki have labelled it as BGA169 but it does not show the extra 16 (typically unused) pins. Count the number of pins (I dare you!) - there's only 153. At any rate, our target SanDisk chip should look like the BGA153 diagram above. Most of the pins are unused / irrelevant for our reading purposes.
The ever helpful GSMhosting site shows us what a full BGA 169 looks like:

BGA-169 Layout - the extra 16 pins comprise the 2 arcs above/below the concentric squares

Other pin arrangements we've seen include BGA162/186 and BGA/eMCP221. Some Flash chips are combined in the same package as the RAM. These are called eMCP (Multi-Chip Package).
Control-F Digital Forensics have blogged an example list which matches some common devices with their e.MMC pin arrangement/size. They also note that the pitch (spacing between pins) for the previously mentioned layouts is 0.5 mm.

So here's what BGA-162 looks like:

BGA-162 Layout (Source:

And a BGA/eMCP221 looks like:

BGA/e.MCP221 Layout (Source:

Final Thoughts

Due to e.MMC standardisation, reading the data off an e.MMC Flash chip should be straight forward and repeatable - which is great for forensics. Interpreting the subsequent data dump artifacts is usually a more challenging task.
The e.MMC Flash memory content discussed in this post applies equally to Smartphones, Tablets etc.

UPDATE: For even more details on Flash Memory Forensics, check out the following papers:
Forensic Data Recovery from Flash Memory

By Marcel Breeuwsma, Martien de Jongh, Coert Klaver, Ronald van der Knijff and Mark Roeloffs


Theory and practice of flash memory mobile forensics (2009)
By Salvatore Fiorillo
Edith Cowan University, Western Australia

The paper by Breeuwsma et al. is probably THE paper on Flash memory Forensics.

Please don't forget to click on this link so Monkey can get his Precious Amazon Echo. You might like to do it from a VM if you're worried about security.
If, for whatever reason, monkey doesn't get an Echo - it's no big deal. Just thought it would make for an interesting exercise as we head towards the Internet of Lazy Fatties ... At the very least, we have learnt more about performing e.MMC Flash memory forensics.

In other news, in June 2016, this monkey will be:
- Attending his first SANS DFIR Summit
- Speaking on a "Innovation in Mobile Forensics" panel with Cindy Murphy, Heather Mahalik , Andrew Hoog and Chris Crowley. Monkey is still pinching himself about joining the collective brain power of that panel *GULP*
- Facilitating/Rockin' the Red Apron for SANS FOR585 Advanced Smartphone Forensics with Cindy Murphy (just after the DFIR Summit)

So if you see me around (probably hiding behind/near Cindy or Mari DeGrazia), feel free to say hello and let us know if this blog site has helped you ... I promise I'll try not to fling too much shit (while you're facing me anyway. Hint: Keep eye contact at all times!).

As always, please feel free to leave feedback regarding this post in the comments section below.