Title: User’s Guide for PRISM Skype Collection
Release Date: 2014-12-28
Document Date: 2012-08-01
Description: This NSA document from August 2012 provides guidance for analysts on ordering the PRISM collection of Skype communications, capabilities that were added over the course of 2011: see the Der Spiegel story Prying Eyes: Inside the NSA’s War on Internet Security, 28 December 2014.
Document: TOP SECRET//COMINT//NOFORN
(TS//SI//NF) User’s Guide For PRISM Skype Collection
POC For Document:
l. (TS//SI//NF) Introduction
a. PRISM Skype collection can be identified by the following:
a.i. SIGAD: US-984XN (this SIGAD is used for all PRISM collection, not just
a.ii. Case Notation: P7*, where the * represents a wildcard.
b. Sustained Skype collection began in Feb 2011 against “in and out” modes; this is
the mode in which one end of the call is using a Skype application, and the other
end is using a landline, or cell phone, without a Skype application. This is audio
only. PRISM collection added exploitation against peer-to-peer Skype applications
in July 2011. In this mode, both/all users are using the Skype application. This can
be a mixture of audio, video, chat, and file transfers. PRISM collection can be from
both modes of communication; however, there is no distinction made in the Case
c. PRISM collection is available for “surveillance” tasking in UTT. There is no “stored
comms” or “search” capability for tasking.
2. (TS//SI//NF) Questions on tasking Skype Selectors to PRISM:
a. Can I task something other than SkypeUser?
a. i. No - this is your only option for Skype selectors in PRISM. You cannot task
b. Do I need to include a V-Series zipcode?
b. i. Yes. When you submit FAA Skype requirements, please remember to include
a V-series zipcode - otherwise, voice content will not be routed to
Derived From: NSA/CSSM 1-52
Declassify On: 20361001
c. Can a Skype user have more than one skypeMailToken?
c.i. Yes - they can link 2 or more email addresses to their account... each has a
corresponding skypeMailToken. You can query DECODEORDAIN to find out
what e-mail address they correlate to.
d. Is there any benefit (or harm) to tasking both the SkyUser and SkypeHash for
the same user to PRISM?
d.i. Tasking a SkypeUser automatically tasks the SkypeHash to UTT. The only
time you need to explicitly task a SkypeHash is when you don't know what
the correlated SkypeUser is (i.e. querying DECODEORDAIN with the
SkypeHash fails to return a correlated SkypeUser, and after you've contacted
ind they fail to find a correlated SkypeUser).
3. (TS//SI//NF) Questions on locating Skype PRISM data:
a. How do I find Skype data in PINWALE?
a. i. Best way to query in PINWALE is by DecodeOrdain enriching your target's
Skype username, as follows:
a.i.l. Enter the Skype username in the PINWALE query form, right
click on the username that you've typed-in and click "DecodeOrdain"
a.i.2. Pick "SkypeUser" from the "choose realm" pull-down list and
a.i.3. You'll now see a screen like the one attached. Pick which ever
permutations you want (or just leave them all checked to get
everything), and click "OK"
a.i.4. submit your search
b. How do I locate Skype collection using DECODEORDAIN?
b. i. Attached is a screen-shot of how you can use DECODEORDAIN to find
Skype video & chat content.
b.i.l. By selecting the "General Skype" permutation, you will query all
Skype PRISM traffic (regardless of whether there is any content or not
b.i.2. By selecting only the "Skype Chat" permutation, you will narrow
the query to only Skype PRISM traffic that contains Instant Messaging
b.i.3. By selecting only the "Skype Webcam" permutation, you will
narrow the query to only Skype PRISM traffic that contains at least
one visible video stream
b. ii. Please note that when MARINA says that a SkypeUser webcammed with
another user... that event *did* actually happen - but the content may not be
visible in PINWALE (because part of the video stream may have been lost,
prohibiting us from being able to reconstruct the video). If you want to look
for PINWALE documents that have webcams that are actually visible, you
can use the DECODEORDAIN-enabled PINWALE query option, described
c. My Skype PINWALE cut says "PRINTAURA Skype" - what does that mean?
c. i. The display name is “PRINTAURA Skype” for both the UIS Text Presenter
and the DNI Presenter. The DNI Presenter and UIS Text Presenter both
have the capability to display SKYPE; Chat, File Transfers, and Video. The
DNI Presenter has a single combined display for SKYPE webcam that allows
the user to play the video and audio together in a synchronized mode or
individually. Within UIS, the user will need to use UIS Hotzone to play the
corresponding audio cuts.
d. How do I search for Skype data in NUCLEON?
d.i. You can query by case notation (P7* where * = wildcard), or you can query
by username in the TELEPHONE_NUMBERS field (note that if your target
has a long username, you may have trouble searching by username due to
length limitations in NUCLEON... but you won't have any trouble searching
by case notation in NUCLEON).
d.ii. Another way to pull up the audio in NUCLEON is to pull on the zipcode that
the audio was sent to. If you pull it up this way, the Skype audio displays
along with the rest of the FAA audio that was sent to the same zipcode.
e. Where's the other side of my voice cut in NUCLEON?
e. i. CES has added a great new service for OPIs on PRISM (and FBI FISA)
voice collection: all voice cuts will now be autopaired and you will not have to
hunt down the other side of phone calls. You will automatically be presented
with both sides of the conversation in NUCLEON. Previously, analysts would
have to search through NUCLEON to find both sides by looking at the
targets and the timestamps to guess which ones are paired.
f. Where’s the audio to go with my Skype video?
f. i. Within the DNI Presenter (DNIP), the user can utilize the “View Associations”
service to find the associated NUCLEON audio of the PINWALE document
or find the associated PINWALE document of the NUCLEON audio. In other
words, you can find the MAM and/or TAM associations to any associated
Skype data, and then display them within the DNIP Skype combined display
or the DNIP Composite display. Also from the UIS Text Presenter, you can
launch the DNI Presenter to utilize this “ViewAssociations” service.
g. How Do I Tell When Stored/Search Comms Arrive From PRISM?
g. i. There is no collection capability for Stored/Search Comms in Skype PRISM.
h. Why do I receive multiple copies of Skype chat sessions?
h. i. You might get chats in segments and then get the whole chat in a third
collect. This is how Skype works. Depending upon what your target is doing,
a copy of his chat history can be sent in-bulk (which can span multiple chat
sessions). If you target, for example, has 3 separate chat sessions with
another individual on his laptop, then logs-into his Skype account on his
desktop, the chat-history of those 3 separate chat sessions will be
transmitted from this laptop to his desktop so that both his computers have a
log of the whole conversation.
/', What does “0 Frames Available" mean in my Skype video results?
i.i. Whenever you see this in PINWALE: “Stream 1 has 0 frames available” or
“Stream 2 has 0 frames available”, it means that we know that a video-feed
was taking place, but we didn't get enough traffic to reconstruct any video
frames. Either the target was unable to get a video feed up, or too much
traffic was lost during the collection process to reconstruct any images.
j. Why are so many e-mail addresses correlated to one Skype user account?
j. i. These are probably all valid associations. Skype does not require that a user enter
a "valid" email address when creating a Skype account (which establishes their
skypeMailTokens). A simple email address like "email@example.com" is likely to be
randomly picked by a large population of Skype users when creating their
accounts... so it's not unreasonable to find a lot of correlations. You will likely find a
similarly high correlation count for something like "firstname.lastname@example.org" or
"email@example.com" or "firstname.lastname@example.org". These are not false hits... these users
*did* enter this email address when creating their Skype account... but a lot of them
probably did so to maintain their anonymity by typing in something simple &
random, rather than their true email address.
k. What's wrong with the timestamps in my Skype data?
k. i. This is due to 2 reasons:
1) Skype sends chunks of old conversations to synchronize conversation
histories across multiple Skype installations. DNI Presenter & UIS attempt to
detect "historic conversation dumps" and label them as such (but keep in
mind these conversation histories can be quite old... days or weeks - so it's
important to pay close attention to the timestamps).
2) There are 2 timestamps displayed in UIS & DNI Presenter for Skype. The
first is the collection time, and the second is the target's local computer time.
DNI Presenter displays them properly as "Collection Time (GMT)" and "User
workstation time (GMT)". Please note that NSA has no control over the "User
workstation time"... a target could decide to set his local computer's clock
months or years off from the true time.
/. IMiy do I sometimes have to piece together a lot of chat sessions from Skype?
l.i. The reason why you're seeing pieces of Skype chats in different documents is for 3
1) the data is too big, and is being fragmented into multiple documents (this isn't
that common, however)
2) Skype sends chunks of old conversations to synchronize conversation histories
across multiple Skype installations. These old conversations are sent piecemeal
across multiple chat sessions.
3) Every time a Skype user disconnects from Skype, and reconnects again - they
being a new session. It's common to see Skype users disconnect & reconnect
frequently whenever Skype stops performing well for them (i.e. if their video or
audio messes up, it's common to see a Skype user disconnect & reconnect to get it
working again). Every time they do this, a new PINWALE document is started.
Items 2 and 3 are based entirely on target behavior (it's not a collection problem,
and is not being introduced by NSA).
m. Why does my Skype playback get out of sync?
m.i. The main issue is data-loss causing the audio to skip and fall-behind. CES is
working on a few techniques to help keep the audio in-sync between channels and
with the video (the video should always be in-synch, since we timestamp each video
4. (TS//SI//NF) Questions on miscellaneous issues:
a. How do I find out the e-mail address of my Skype target? The collection only
gives me the Skypeuser, SkypeNode and SkypeMailToken.
a. i. Email addresses are not included in Skype collection... skypeMailTokens
(which are derived from email addresses) are. You can search MARINA for
SkypeMailToken - to - SkypeUser correlations to detect a username change,
in order to put new SkypeUser selectors up on FAA coverage. You can also
transform any email address into a SkypeMailToken using
DECODEORDAIN. DECODEORDAIN can also transform some
skypeMailTokens back into email addresses. For cases where
DECODEORDAIN cannot transform a SkypeMailToken into an email address
- please send an email to^^^^^^^^^^and request that they attempt
to recover the email address from the SkypeMailToken (however, please only
do this if DECODEORDAIN fails to return the email address).
b. Can we collect Skype voicemail?
b. i. If a target connects via a peer-to-peer connection, we will not be able to
collect any voicemail he receives or sends. If a target connects via a PSTN
connection and uses his In/Out number to communicate with another user’s
In/Out number, then we will collect voicemails. Both of these scenarios only
apply to surveillance collection. We do not have any stored comms collection
from Skype at this time.
b.ii. I/O is a telephone number assigned by Skype. You can use either your
Skype ID or your I/O number to make/receive calls. If you’re only making
P2P calls you don't need an I/O assigned. You only need an I/O number if
you want to make calls to/from the PSTN. Note you can also make I/O to I/O
calls, which must hit a PSTN switch of some type.
c. What happens when both sides of a Skype call are tasked selectors?
c.i. PRISM should collect two copies of the communication. This is true unless
one of the users' Skype communications falls into one of a couple of
technical limitations which prevents us from currently collecting via PRISM.
d. Why did I receive small and incomplete files from Skype collection?
d.i. The first 256 bytes of a document for transfer are sent so that the Skype
client can figure out what type of file it is (a PDF verses a JPEG). It shows-up
on the user's Skype chat window as a message that says something like "so
and so is attempting to sent you a file... click here to download... etc.". If the
user doesn't choose to download it, then you will not receive the entire file,
but only this first nibble of data, which might include the filename.
d.ii. Also, due to known limitations in our collection process, some larger size
files being transferred by targets may be not be fully collected. There is no
stored comms capability to go back and retrieve these files.
e. How does the Skype Credential CRC work and what will I see in collection?
e.i. The way Skype works, users (in the raw data) are represented by their
Skype Credential CRC (SkypeCredCRC). Usually the data contains enough
information such that we can figure out what SkypeUser maps to the
SkypeCredCRC... but sometimes this isn't possible. SkypeCredCRC's are
temporary identifiers... so they're not targetable. Normally analysts don't
have to worry about SkypeCredCRC's, because CES Protocol Exploitation is
able to map them to SkypeUser's... but there are a handful of edge cases
where they don't have enough information to do so. Also, a SkypeUser can
have more than 1 SkypeCredCRC associated with them. So - the possibility
that a SkypeCredCRC shows-up in your collection and is associated with
one of the already-existing communicants in the collection is a very possible
(although it's not universally true). CES only "holds-up" collection for a
certain period of time before they give-up trying to figure-out how to map an
unknown SkypeCredCRC to a SkypeUser (12 hours would be too long... that
would build-in a 12-hour delay into all Skype collection before you got to see
it). If you rurnnt^^SkypeCredCR^an^annot figure out who it is, just send
an email to and someone (either in PE, or
SSO) can try and figure-out if we (or the FBI) has learned the correlation
after-the-fact. It is not possible for you to see the correlation in the raw data.
Encryption is used to mask the correlation.
f. How does Skype work?
f.i. Skype is a true "peer to peer" service... communications are not routed
between central servers - they are only routed between Skype users (i.e.
"peers"). NSA receives Skype collection via PRISM when one of the peers is
a FAA/702 tasked target.
There are 3 collection scenarios:
A) target --> other_person
B) other_person -> target
C) target -> same_target
Scenario A and B occur when a target is communicating with another
individual. This communication may not be very interesting (i.e. it might not
be an IM, Voice, or Video... it may just be a request to retrieve that
individual's Avatar to display)... but there was *some* type communication
between these two users, and the target is a party to that communication.
Scenario C is an interesting case -- it implies that the target is talking to
himself. In fact, he is... This scenario occurs when the target uses 2 different
devices (i.e. laptop and mobile phone, or laptop & desktop, etc.), and is
logged in from both devices at the same time.
These 2 devices will automatically "synch" with each other, and transmit
information between them (including past communications that the target
was a party to).
Now - when it comes to multi-parti communications (conference calls, multi-
person IMs, etc.), Skype is still a peer-to-peer protocol (1-to-l). Skype
creates a mesh-network, where users are connected together through
multiple peer-to-peer links. Instant Messages sent to this group of meshed
participants can be routed through* any participant. For example, if A, B,
and C are chatting together in a conference chat, the peer-to-peer network
may look like this: A B C. In this scenario, IM's from A to B and C
will all be routed through B. If B is our target, we will see these IM's... but B is
a party to these IMs - because they're multi-party IMs (not private IMs
between A and C).
5. (ts//si//nf) Other Useful Resources:
a. Instructional Videos on KapShare Learning Portal
a.i. Go to this link:
and type in “Skype” in the Search box. You will see a list of short videos
which describe different aspects of Skype exploitation. As of Nov 2011, the
list includes a primer on how Skype works, how to analyze Skype metadata,
Skype in XKeyscore and other topics.
b. Email the PRISM technical help team with any Skype question:
c. Visit the PRISMFAAweb site and search the FAQ section for Skype answers and