LAS VEGAS – OCTOBER 31, 2001 Pluvia Securities Research INITIATES coverage of Viisage Technology, Inc. (NASDAQ: VISG), at $14.25, with a STRONG SELL Recommendation and a target price of under $2.00 per share.
OUR STRONG SELL RECOMMENDATION IS BASED ON:
1. A recent study of facial recognition software by the United States Department Defense that demonstrated all currently available facial recognition software, including Viisage’s product, failed to identify would be terrorist subjects, under ideal conditions, more than 75% of the time at a distance of 3 meters, while falsely identifying would be “innocent people” as criminals, at rates unacceptable in an airport security installation. The results of this test prove other existing security technologies provide more accurate methods of identifying terrorists in airports.
2. Viisage’s long history of issuing fraudulent and misleading statements and press releases which overstate the effectiveness of the Viisage Facial Recognition system; misrepresent Viisage’s relationship with other significant companies; and are currently intended to capitalize on the tragic events of 9-11 by promoting fraud;
3. Violation of SEC disclosure rules that require Viisage to report significant events including the DOD study results that suggest Viisage’s facial recognition product has limited viability as a commercial security tool;
1. DOD STUDY DEMONSTRATES HIGH FAILURE RATE
· In February 2001, the CDTDPO office of the Department of Defense released results of the most extensive test of automatic facial recognition systems ever undertaken, titled “Facial Recognition Vendor Test (FRVT) 2000”. Viisage’s system was included in this test, although Viisage has failed to disclose these results to shareholders.
· The following results of the “Distance Experiments” conducted in this test are an excellent example of the performance you can expect from a facial recognition system used for surveillance in an airport to identify terrorists.
[The term “gallery” image means an enrollment image in a database. The term “probe” image means a test image to be compared to each image in the “gallery”. ]
“Distance” Experiments. The first such experiment is really a test of a high quality gallery collected on a digital camera at 1.5 meters against probe images collected with a lower quality video camera at 2.0 meters. There are about 185 images in each set, taken on the same day and under controlled lighting conditions with cooperative subjects. The best vendors correctly matched faces 40% of the time, while falsely matching only one in a hundred.
The second distance test was identical to the first, but placed subjects 3 meters away in probe images. Performance for all vendors dropped considerably, the best correctly matching faces almost 25% of the time while falsely matching only one in one hundred.
The third distance test moved the probe images to 5 meters. Faces were correctly matched 15% of the time while falsely matching only one in one hundred. http://www.dss.state.ct.us/digital/news24/bhsug24.htm
Note that these results were achieved under ideal conditions. According to the DOD study - actual results will likely be worse. In real life conditions, the source image most likely won’t be high a quality, digital format, taken in optimal lighting conditions; photos will certainly not be taken on the same day, therefore aging will cause worse results; facial hair and glasses may be different; angles of the photos cannot be assumed to be the ideal.
The DOD study examined each of these issues and found every one caused a significant drop in the software’s ability to correctly match a subject with a database picture, as noted in these DOD results from the “Illumination Experiments”:
“Three illumination experiments were performed. In all experiments, the gallery images were high quality mug shots. In the first experiment, the probe images were taken with overhead lighting; in the second, with “badge system” lighting; and the third with outdoor lighting. There was no template aging. Outdoor lighting caused the greatest decrease in performance, with systems performing at the 40% correct recognition level with one false match out of a thousand” http://www.dss.state.ct.us/digital/news24/bhsug24.htm
2. HISTORY OF FRAUDULENT PROMOTION
FRAUDULENT PROMOTION #1 - Taking advantage of the 9/11 tragedy
Following the events of 9/11 Viisage capitalized on the tragedy by falsely suggesting Viisage’s Facial Recognition system might have identified the terrorists had it been deployed on 9/11, and falsely stating Vissage’s software was “99+% accurate”. Several examples of these intentionally misleading claims include:
· INTENTIONALLY MISLEADING: Oct 7 article in the Atlanta Journal regarding the 9/11 terrorist attacks: "This technology could have made a real difference had it been deployed"; says Tom Colatosti, the chief executive officer of Viisage. http://www.accessatlanta.com/ajc/terrorism/opinion/1007toner.html
· INTENTIONALLY MISLEADING: Oct 17, Matt Lauer, NBC Today show interview regarding 9/11: “if this technology had been deployed at the airports, there--it could have made a significant difference in terms of helping prevent it.”
· INTENTIONALLY MISLEADING: Sept 17 Viisage Press Release purposely titled to promote VISG stock in the wake of the 9/11 disaster: “Viisage Technology Offers FBI Free Use of Face Recognition Technology to Aid Investigation”. Viisage’s Colatosti claims: “It is clear that the use of face-recognition technology can make a real difference in identifying terrorists.” http://biz.yahoo.com/bw/010917/172450_1.html
· INTENTIONALLY MISLEADING: Fortune article: "We've had a case where a person aged ten years, gained30 pounds, and grew a beard, and we were still able to identify him,’ says CEO Tom Colatosti, who claims his software is 99.7% accurate. http://www.fortune.com/indexw.jhtml?channel=artcol.jhtml&doc_id=204613
· INTENTIONALLY MISLEADING: Oct 8 Dow Jones Interview: Colatosti said he believes every airport in the country will want to install the system which scans people's faces and compares them to stored images in a database. …”We might not get the contracts for every airport, but we'll get our fair share" Colatosti said, adding,” Maybe we might even get our unfair share." quicken.com
· THE TRUTH – 1. DOD STUDY PROVES VIISAGE SYSTEM IS NOT 99% ACCURATE
· The Department of Defense study quoted herein is the most extensive test of automatic facial recognition ever undertaken. Dr. Jim Wayman, director of the National Biometric Test Center at San Jose State University is an expert on Biometrics and facial recognition. Dr. Wayman concluded his review of this study by writing:
”If we compare each person to a database of 100,000 and wish to narrow the selection to 1,000 potential matches, we would fail to detect a real fraudster 40% of the time. This seems to throw into question some of the recently announced projects for use of facial recognition to prevent multiple enrollments in large databases.” dss.state.ct.us
· In a September 27 WSJ article Dr. Wayman is quoted saying the DOD study found “the very best of the [facial recognition] systems failed a third of the time”.
· The DOD test results prove Viisage’s facial recognition system falls well short of being“99% accurate” as Viisage CEO Colatosti has claimed numerous times to national audiences since 9-11. These intentionally misleading statements are particularly disturbing when you consider the likelihood they were made in an attempt to artificially inflate the value of VISG stock and fraudulently capitalize on the tragic events of 9/11.
· As a public company Viisage is required by law to disclose the truth about the performance of their facial recognition software. Each quote attributed to Viisage above illustrates an incident where Viisage has intentionally mislead the public in direct violation of securites law.
THE TRUTH – 2. FACIAL RECOGNITION WOULD NOT HAVE CAUGHT THE 9-11 TERRORISTS
· On 9/11, pictures of only 2 of the 19 terrorist were available in FBI computers. Without pictures of the terrorists, Viisage’s technology would not have detected them.
· Terrorists took test flights prior to 9-11 and would have known if they could not pass any airport security measure.
· Terrorist organizations have recruited hundreds of individuals from Egypt, Saudi Arabia, Yemen, Sudan, Somalia, Pakistan, Afghanistan and elsewhere. There is no chance whatsoever that any existing photo database would hold even a significant percentage of these individuals at any given time, much less all of them.
THE TRUTH - 3. NOT EFFECTIVE FOR CATCHING TERRORISTS IN AIRPORTS
· Results of the DOD study conducted under ideal conditions prove failure rates of facial recognition software render the technology useless in surveillance situations such as airports - where prevailing conditions would be much more demanding.
· Security and facial recognition experts agree that the large number of false matches, or “cry wolf failures” would effectively overwhelm security personnel, conditioning them to assume all matches were mistakes.
More Effective Security Technology Available
· When comparing the failure rates of other technologies currently used in airports with the failure rates of facial recognition systems, it becomes obvious very quickly that other existing technology – such as INSPASS are considerably more effective due to higher accuracy rates and lower false positive rates.
· INSPASS systems for example allow passengers to use a smart card with biometric links to their fingerprint or hand. When passing through security, they swipe the card, place their hand on a reader and are identified. ins.usdoj.gov
FRAUDULENT PROMOTION #2 - “SNOOPERBOWL 2001”
During last year’s Super Bowl in Tampa Bay FL, Viisage and the Tampa Bay Police department used Viisage’s facial recognition system in a test to compare faces of fans that attended the game with criminal suspects from a police database.
In numerous interviews after the “Snooperbowl“ event, Viisage’s CEO Tom Colatosti intentionally misrepresented both the results of the test and the capabilities of the Facial Recognition system when he claimed the test was a “great success” and that “19 arrests were made”. Colatosti also made numerous references to 19 “hits” achieved by their system suggesting the “hits” had identified criminals:
· FRAUD - September 17 VISG PR: “Our technology was used with great success at SuperBowl XXXV, where the security concerns were similar"; biz.yahoo.com
· FRAUD - “Most notably, our face technology was used at the Super Bowl XXXV, one month ago in Tampa…” “…It was publicly reported that we obtained 19 image matches. This was a significant technological feat and has been widely acknowledged as being successful by the law enforcement community.” viisage.com
· FRAUD - In an interview on the WFLA-AM 970's morning show Colatosti jubilantly promoted the” great success” of the Facial Recognition system at Snooperbowl claiming it had resulted in “19 arrests” pqasb.pqarchiver.com
THE TRUTH – NO ARRESTS, NO PROOF OF CRIMINALS
· Contrary to Viisage’s claims, during the Snooperbowl test there were no “arrests”. Furthermore, there were no stops to confirm the identity of the persons recognized by the Viisage system as “criminals”. Without stopping the individual and checking their identity, there is no way of knowing if persons recognized by the Viisage system were identified correctly or simply a “false match”. Contrary to Viisage’s claims, there was no evidence whatsoever supporting the claim that the Viisage system correctly identified a single criminal.
· When the Viisage system generated a "hit" at the Superbowl, it showed side-by-side pictures of the person identified as a criminal at the turnstile with a picture from the database. Officer Bill Todd of the Tampa Bay Police Department who worked the “Snooperbowl” test stated: "I looked at some of those side-by-side pictures, and they weren’t of the same person".
· DOD study results of Viisage’s facial recognition software proved the Viisage system would have statistically produced at least 19 false “hits” which would have had the effect of incorrectly identifying innocent persons as criminals in the “Snooperbowl” test. dodcounterdrug.com
FRAUDULENT PROMOTION #3 – MISREPRESENTED RELATIONSHIPS WITH SIGNIFICANT COMPANIES
· HYPED NEWS RELESES “Contrary to a Viisage news release dated March 17, 2000, Lockheed Martin has not entered into a technology partnership to pursue biometrics opportunities.”
lockheedmartin.com
3. FAILURE TO DISCLOSE MATERIAL EVENTS
SEC regulations require publicly traded companies to disclose significant events. The high failure rate shown in the DOD study calls into question the commercial viability of the Viisage facial recognition software. This is clearly a “material event” and by law Viisage should have disclosed the results to shareholders.
RESULTS OF INSTALLED FACIAL RECOGNITION TESTS
· In Massachusetts, state welfare officials have suspended their use of the Viisage face-recognition system after four years of free use, 15 matches and no arrests. They say they aren't sure if it's worth it.
· The U.S. Immigration and Naturalization Service experimented with face recognition at the Mexican border but discontinued the test. An INS spokesman confirmed that the service had tried a facial-recognition system at the border, but stopped working with it, stating the technology would work if everything was just right, but that's never the case.
· The Tampa Police Department installed a facial recognition system in Ybor city on June 29, 2001. 4 months later officer Todd, who is involved with the test, admits that so far there had been no face matches and no arrests in what he described as "a test with great expectations."
· West Virginia officials are now using the software to search for children "24 hours a day, seven days a week," so far without success, according to an official involved in the project.
· Capt. Dan McCoy of the Santa Ana, Calif., police department, which has had a face-recognition system for two years, said it was useless at making identifications from grainy store-surveillance videos.
· The biggest face-recognition installation in public space is in Newham, England, a rough neighborhood in London's East End. In this test 300 security cameras are trained on the five main shopping areas. After several years of service this face recognition project has not led to a single arrest.
For these reasons, Pluvia Securities Research issues a Strong Sell Recommendation on VISG with a 6-month price target of under $2. Pluvia Securities Research will maintain research coverage of Viisage and update future findings.
Pluvia Securities Research, their agents, associates, and or employees have investment positions consistent with the above-stated investment opinion. All comments are the EXPRESS opinion of the Author(s) - All rights reserved.
CONTACT:
Pluvia Securities Research
Steve Pluvia |
