Al-Zawahiri: At It Again!

Open PDF
Creating agency: NSA National Security Agency (US)
Document date: 2006-03-17
Type: SIDtoday
Pages: 2
Genre: newsletter
Banner:
Classification: SECRET SI
Dissemination: -

Analysis of a new video from Al Qaeda leader Ayman al-Zawahiri shows near-perfect voice matching. The NSA is exploring a hand-held voice recognizer for use in “detention centers and guard posts in conflict regions.”

Long Description:
This SIDtoday article analyzes a video of Al-Qa'ida deputy Ayman al-Zawahiri released by Al-Jazeera on 5 March 2006, comparing his voice mathematically against previous transmissions to achieve near-perfect voice matches using a new capability from MIT's Lincoln Laboratory. The author describes how these voice-matching technologies, combined with miniaturized computing, could be integrated into a hand-held device to identify persons of interest in detention centers and guard posts in conflict regions. The document also covers digital and video analysis of the taping, including collaboration with US Secret Service analysts to check for concealed details in the video background.
Programs:
-

People mentioned:
Cusmariu, Adolf, al-Zawahiri, Ayman, bin Laden, Osama

Countries:
-

Tags:
speaker identification, Al-Qaeda, video analysis, biometrics, counterterrorism, detention

Identifier:
2006-03-17_SIDToday_-_Al-Zawahiri_At_It_Again

PDF SHA-256:
0b12974d2f94dba122ad79178367e3afc489eca1b87a73ab2d2cce0c11061593

DYNAMIC PAGE -- HIGHEST POSSIBLE CLASSIFICATION IS
TOP SECRET // SI / TK // REL TO USA AUS CAN GBR NZL

(U) Al-Zawahiri: At It Again!

FROM: Adolf Cusmariu
Technical Director, Content Analysis Services (S21212)
Run Date: 03/17/2006

(S) What does an analysis of the most recent video from Usama bin Laden's deputy tell us?

(U) In a transmission released by the Arab Al-Jazeera broadcaster on Sunday, 5 March 2006, Al-
Qa'ida number two Ayman Al-Zawhiri offered yet another diatribe against the US and its allies.
This production was decidedly civilian; the terrorist is shown in a home setting without his usual
Kalashnikov rifle (see image below); the significance of the black turban - as opposed to a white
one worn in some previous tapings - is unclear.

(S//SI) Unwittingly, the long 7.5 minute message offered ample opportunity to check
mathematically for a voice match with previous transmissions, which proved mostly perfect. The
five dates compared since the summer of 2005 -- 6/17/05, 9/1/05, 1/6/06, 1/30/06, and
3/5/06 -- yielded high-confidence (99%) pairing estimates, except (surprisingly) between the
September/March transmissions.

(S//SI) However, a new capability recently delivered to NSA by MIT's Lincoln Laboratory allows
optimal combination of vocal-tract models from contentious intercepts. The combined
September/March vocal model matched perfectly the other dates; thus, careful modeling for
mathematical comparison is a critical step in voice matching , to ensure fewer false
alarms and increase the probability of detection. This will become critically important when
large-scale voice processing capabilities are deployed in-theater , attempting to identify
persons bent on terrorist activities against US forces or the local populace.

(S//SI) Now, these technologies, together with miniaturization in computing power, could be
integrated into a hand-held device for field application in hostile environments. Such a
device is pictured below, which would store in its memory voice models of Persons-Of-Interest
(POI's), then compare an acoustical sample from a suspect with the POI's, producing match
estimates from MIT's mathematical engine. Detention centers and guard posts in conflict regions
needing authentication or cross-check capabilities would obviously benefit from this advanced
component in security operations; fabrication options are being explored.

(S) Returning to the latest Zawahiri taping, it bore all the imprints of a home video; digital
comparison with the previous transmission of 1/30/06 shows the difference (see below). Image
sections of 100 x 100 pixels covering an area around the speaker's left shoulder clearly display
(in gray scale) the jagged edge nature of the 5 March video, suggesting increased compression
rates for possible web relay.

(S) Finally, additional video analysis was performed on the controversial 1/30/06 transmission
(see related article ) - with help from our colleagues at the US Secret Service (analyst [REDACTED]
[REDACTED]) - to determine if any unusual patterns or structures were present in the black
background. An example of video frame integration is shown below:

(S) No concealed details of any sort were obvious, either in this image or any of the
several other operations performed on the color video. However, the background - most likely a
black curtain - does show some unexplained shadows, perhaps due to lighting or persons
walking behind the curtain during taping.

(U//FOUO) On a personal note, US Secret Service management and the half-dozen image
analysts gathered around, intrigued by this problem, found the sense of collaboration with NSA a
privilege.

"(U//FOUO) SIDtoday articles may not be republished or reposted outside NSANet
without the consent of S0121 (DL sid_comms)."

DYNAMIC PAGE -- HIGHEST POSSIBLE CLASSIFICATION IS
TOP SECRET // SI / TK // REL TO USA AUS CAN GBR NZL
DERIVED FROM: NSA/CSSM 1-52, DATED 08 JAN 2007 DECLASSIFY ON: 20320108