Another Sunday Times story about some alleged technology which is being hyped to the surveillance state snoopers, and about which some critical questions need to be answered.
The Sunday Times November 26, 2006
Word on the street ... they're listening
POLICE and councils are considering monitoring conversations in the street using high-powered microphones attached to CCTV cameras, write Steven Swinford and Nicola Smith.
The microphones can detect conversations 100 yards away and record aggressive exchanges before they become violent.
The devices are used at 300 sites in Holland and police, councils and transport officials in London have shown an interest in installing them before the 2012 Olympics.
The interest in the equipment comes amid growing concern that Britain is becoming a "surveillance society". It was recently highlighted that there are more than 4.2m CCTV cameras, with the average person being filmed more than 300 times a day. The addition of microphones would take surveillance into uncharted territory.
These estimated figures date from 2003 - there could be many more than this installed today, but nobody knows for sure, because they are not licensed or regulated.
The Association of Chief Police Officers has warned that a full public debate over the microphones' impact on privacy will be needed before they can be introduced.
The equipment can pick up aggressive tones on the basis of 12 factors, including decibel level, pitch and the speed at which words are spoken. Background noise is filtered out, enabling the camera to focus on specific conversations in public places.
That may be what the company trying to flog the system claims, but there is no evidence that this is any more accurate than a system which merely detects human voices.
If the aggressive behaviour continues, police can intervene before an incident escalates. Privacy laws in Holland limit the recording of sound to short bursts. Derek van der Vorst, director of Sound Intelligence, the company that created the technology, said: "It is technically capable of being live 24 hours a day and recording 24 hours a day. It really depends on the privacy laws in a particular country."
The same moral and ethical stance, as taken by arms dealers, and suppliers of torture equipment.
Last month Martin Nanninga of VCS Observation, the Dutch company marketing the technology, gave a presentation to officials from Transport for London, the Metropolitan police and the City of London police about the CCTV system. Nanninga is to return next year for further discussions.
"There was a lot of interest in our system, especially with security concerns about the Olympic Games in 2012. We told them about both our intelligent control room and the aggression detection system," Nanninga said.
All sorts of security snakeoil merchants and reputable companies will be trying to leech onto the fat London 2012 Olympics "security theatre" budgets. The security budget for the Athens Olympics in 2004, was over a billion pounds sterling..Nowhere near this sum was included in the bid budget for the London 2012 Olympic Games, only about £400 million for actual stadium event security. Given the higher terrorist target profile of London, who believes that even more British taxpayers' money will not be wasted ?
The British company Autonomy installed software for a gunfire and explosion sound detection and triangulation system for the 1200 CCTV cameras and microphones installed for the Athens Olympic Games in 2004 (see below for further comments about these sort of systems).
Their other security snooping software installed to help monitor Greek phone and email communications for keywords, in multiple languages, is a separate issue. The law enforcement interception software installed by telecomms equipment supplier Ericsson for the Athens Olympic Games, led to the ongoing Vodafone Greece scandal, whereby even the mobile phones belonging to the Greek Prime Minister etc. were intercepted, by unknown persons, somehow abusing the very security system that was meant to be a protection.
In Holland more than 300 of the cameras have been fitted in Groningen, Utrecht and Rotterdam. Locations include city centres, benefit offices, jails, and even T-Mobile shops. The sensitivity of the microphones is adjusted to suit the situation.
What possible need is there for a remote control room to be monitoring "benefit offices, jails, and even T-Mobile shops" ?
Are there really no staff in these locations, with access to panic buttons or mobile phones, who could raise the alarm and call the police if necessary ?
Police and local council officials are still assessing their impact on crime,
We are willing to bet, that there will be no statistically valid improvement in either crime prevention or in the public's fear of crime, as a result of these small scale trials.
There will be an increase in the number of people who feel oppressed by this further extension of the surveillance state.
although in an initial six-week trial in Groningen last year the cameras raised 70 genuine alarms, resulting in four arrests.
What a misleading statistic ! This technology is another form of biometrics, with extra technical difficulties due to the attempt to deploy it in the street, rather than under controlled conditions indoors.
How many false alarms were there, i.e. how many false positives ?
How many incidents or arrests were there in the test areas, which were not detected, i.e. how many false negatives were there ?
How many of the alleged 70 incidents were not responded to by the Police, due to other resource commitments ?
How many of those claimed 4 arrests would have happened anyway, because the Police had already been called by the public, or perhaps even alerted by the conventional CCTV system operating mode, i.e. purely visually ?
Harry Hoetjer, head of surveillance at Groningen police headquarters, recalled an incident where the camera had homed in on a gang of four men who were about to attack a passer-by. "We would not normally have detected it as there was no camera directly viewing it," he said.
So, were these the 4 people arrested above ? If so, on what evidence, if the conversation was not actually recorded ?
How many other groups of innocent people did the system home in on, who did not subsequently attack anyone ?
Last Friday a Sunday Times reporter visited the office of Sound Intelligence in Groningen
"Sound IntelligenceSound Intelligence was founded in 2000, based on research at the departments of Biophysics and Artificial Intelligence of Groningen State University (RuG), into the mechanisms underlying human hearing."
to test the system. The reporter stood in the control centre with a view of an empty room on one of a bank of monitors. Van der Vorst entered the room, out of sight of the camera, and began making aggressive noises.
The camera swivelled to film him and an alarm went off in the control room, designed to alert police to a possible incident.
That is hardly a scientific test of a system which is mostly deployed outdoors, is it ?
How could the technology cope with, for example, groups of football supporters, or political demonstrators going to or coming from a sporting event or political demonstration ?
How can it possibly accurately detect "aggressive sounds" uttered by non-European language speakers, e.g. Chinese ?
"The cameras work on the principle that in an aggressive situation the pitch goes up and the words are spoken faster," said van der Vorst. "The voice is not the normal flat tone, but vibrates. It is these subtle changes that our audio cameras can pick up on."
Unless, of, course you happen to be, say, Chinese, where the rising tone inflections and rapid speech, happen to be the normal way in which you speak.
We do not believe that there is any scientific evidence that you can reliably predict if a shouting match in public will, or will not result in physical violence, especially, if, as in this system, there is no attempt to even bother to try to pick up individual words and assign meaning to them.
It is difficult enough to use "voiceprint analysis" to try to match individuals, or to see if they are speaking differently from their own, previously machine learned "normal" speech patterns, for use in a "polygraph lie detector".
We cannot understand how unknown people, are meant to be classified as "aggressive and likely to be about to attack someone", at a distance, in the outdoors.
Public prosecution services can use them in court as evidence. The Dutch privacy board has already given its approval to the system.
That is of little comfort. The Netherlands is ranks alongside Sweden, just behind the United Kingdom as one of the most surveilled countries in the European Union according to Privacy International.
According to a spokesman for Richard Thomas, Britain's information commissioner, sound recorded by the cameras would be treated under British law in the same way as CCTV footage. Under the commissioner's code of practice, audio can be recorded for the detection, prevention of crime and apprehension and prosecution of offenders. It cannot be used for recording private conversations.
The Information Commissioner's code of practice is not legally binding
BP Petrol stations have been snooping on their customers and staff via audio recordings, as well CCTV and more recently Automatic Number Plate Recognition, for several years now.
Graeme Gerrard, chairman of the chief police officers' video and CCTV working group, said: "In the UK this is a new step. Clearly there is somebody or something monitoring people speaking in the street, and before we were to engage in that technology there would be a number of legal obstacles.
"We would need to have a debate as to whether or not this is something the public think would be a reasonable use of the technology. The other issue is around the capacity of the police service to deal with this."
Will ACPO and the Home Office care to debate the issues regarding existing and developing CCTV survellance technologies with us ?
The technically easy bit of this system is the sound triangulation part of it. Even that is prone to major errors in an urban environment, given the complexities of echoes off buildings etc. and the need to analyse the sounds in near real time, for these analyses to be of any use at all.
Unlike the dubious "aggression sound analysis" proposed by this Durtch system, based on research by the University of Groningen, there have been several firearms or gunfire detection systems on the market for over 10 years now.
These have the technically simpler task of filtering and analaysing characteristic loud gunshot sounds, rather than the subtleties of human voices under stress, and trying to triangulate their points of origin, and to despatch a police or military patrol to investigate.
See this National Institute of Justice Research paper published in 1999
which actually analyses a couple of these systems which were installed by commercial companies in Redwood City, California and in Dallas, Texas, as far back as 1996.
The British Army had a similar non-commercial system set up in Belfast in Northern Ireland, even earlier than that.
Detecting the gunfire and triangulating on its source was not a problem, even with the available technology 10 years ago. It is the resultant police (or military) and community responses, which are the main issues, and they are unlikely to have improved since then.
If the British authorities are stupid enough to waste money on this Dutch or any other "microphones in the streets" systems, then they will have to prove to us, that they will somehow not be vulnerable to spoofing attacks.
Gunshot detection systems can discriminate between the sound of a blank round and a live one, being fired.
There does not seem to be any way for the remote location "aggressive voice analysis" scheme to detect genuine human voices, and pre-recorded ones.
Why will terrorists, or bank robbers, or anarchist demonstrators, not use cheap, high fidelity audio recordings to lure away police patrols from their real targets ?
If such patrols do not bother to investigate apparent "audio incidents", then what is the point of wasting our money on the scheme ?
We hope that this inherently flawed technology is not even trialled in the UK, let alone deployed for either the London 2012 Olympic Games, or elsewhere.
If it is deployed, then we predict that the nightmare of a National Voice Print Database will be mooted by the control freaks, once a large number of microphone equipped CCTV spy cameras are linked together.