This system claims to detect suspicious behavior that indicates shoplifting:
Vaak, a Japanese startup, has developed artificial intelligence software that hunts for potential shoplifters, using footage from security cameras for fidgeting, restlessness and other potentially suspicious body language.
The article has no detail or analysis, so we don’t know how well it works. But this kind of thing is surely the future of video surveillance.
Miguel • March 7, 2019 3:14 PM
A shoplifting detection system, in Japan of all countries. How do they train it? With foreign movies?
Tatütata • March 7, 2019 3:15 PM
My snake oil detector went into overload, analogous to its response for polygraphs or so-called voice tremor analysers.
I would expect this contraption to reflect the prejudice it was trained with, and to reliably pick out the black guy in the crown (or in the Japanese context, the Korean guy?)
Or will it enforce some unwritten social norm on how to “appropriately” behave in a store? Keep your buttocks tight together…
They had at least the grace of choosing an appropriately sounding four-letter name.
They’re not the only ones to work on this stuff. This Japanese patent application filed in 2015 by NTT mentions all the appropriate buzzwords, including cloud-based processing. The prior art section mentions a provider called “Saburo Kun”. The automatic translation of the product description states:
How Saburo-kun A works
Automatic sensing of suspicious behavior / behavior in video analysis!
Stop shoplifting at the staff’s voice.When shoplifters try to shoplift, they will make specific movements different from ordinary. When “Saburo-kun A” analyzes the unique movement through the image, when it detects it, it automatically teaches the clerk an alarm signal and promptly tells the sales clerk, so that “it calls” to the suspicious person as a trigger, It leads to deter shoplifting.
I suspect China leads the pack in that area…
Tatütata • March 7, 2019 3:25 PM
Not only JP and CN…
Honeywell International Inc., US2008031491A1:
In an embodiment, a video processor is configured to identify anomalous or abnormal behavior. A hierarchical behavior model based on the features of the complement of the abnormal behavior of interest is developed. For example, if the abnormal behavior is stealing or shoplifting, a model is developed for the actions of normal shopping behavior (i.e., not stealing or not shoplifting). Features are extracted from video data and applied to an artificial intelligence construct such as a dynamic Bayesian network (DBN) to determine if the normal behavior is present in the video data (i.e, the complement of the abnormal behavior). If the DBN indicates that the extracted features depart from the behavior model (the complement of the abnormal behavior), then the presence of the abnormal behavior in the video data may be assumed.
The named inventors live in the Minneapolis area.
Hopeful • March 7, 2019 6:35 PM
Seems like a job for soulless AI
Roboticus • March 7, 2019 10:45 PM
Given the current regulations on loss prevention in stores it would almost certainly just alert a human on exactly who to watch, which is valuable in a store with thousands of customers a day and 2-3 security people. And if it is wrong? LP watches people who aren’t doing anything wrong anyway without anyone noticing.
Winter • March 8, 2019 2:44 AM
I think building such a system is pretty straightforward.
With just the positions-by-time path and some crude eye/head tracking it must be pretty easy to collect really masses of “normal” data. Look for outliers in two aspects: hesitation in presence of others and looking away from the merchandise. That would already allow you to mark suspicious clients.
With current face detection software and DNN’s, this sounds like a student homework assignment. Getting the false positive rate down to lower single digits might be more difficult, though.
Tony • March 8, 2019 3:08 AM
US TV Show “Person of Interest” main non-human character is an IA, The Machine, acting via IoT and monitoring networks, which can recognize some people’s bad intentions through body language, keywords and so on.
See https://personofinterest.fandom.com/wiki/The_Machine (warning : only if you won’t plan to watch it, spoilers there).
Interesting TV Show about security, privacy, IA.
Alejandro • March 8, 2019 3:41 AM
Even after reading the article I don’t see how this works.
You load up some software to the (store?) computer and it catches shoplifters. Seems like there is a lot left out.
Are we assuming massive deployment of cameras, microphones, foot pressure switches, alarmed merchandise, rfd tracking, facial tracking…or what?
Also, for the software to exist, there must be some well founded behavioral studies around to back it up. I did’t look for that, but would guess anything available is the common sense stuff like, “look for people who make furtive movements”, etc.
I think prospective customers might want to let this technology mature a bit before laying money on the table for it.
DrMcCoy • March 8, 2019 7:09 AM
“fidgeting, restlessness and other potentially suspicious body language”
So I guess this will again hit people with disabilities and mental illnesses like anxieties and ADHD the worst?
Snarki, child of Loki • March 8, 2019 7:57 AM
“So I guess this will again hit people with disabilities and mental illnesses like anxieties and ADHD the worst?”
Plus people trying to figure out where the store hid the restrooms.
maddox • March 8, 2019 9:52 AM
Miguel, oddly, it’s becoming somewhat common for elderly Japanese to shoplift, in the hopes that they’ll be caught and go to prison. Of course, training with videos of people trying to get caught might not be ideal.
Realistically, only stores with shoplifting problems will want this product, and they’ll be able to provide videos to improve it.
Tatütata • March 8, 2019 10:25 AM
The comment of 9:52AM suddenly reminded me of a 2018 Japanese film I recently saw called “The Shoplifters“. Its subject is really about answering the provocative question “what is a family?”, but the film includes several shoplifting scenes. The group of people without shared DNA it portrayed was way more functional than the prototypical conventional nuclear familyn, but wouldn’t fit in the straight jacketed society as represented by its police, judges, social workers, etc., which ultimately come across as tools of repression.
The viewers’ opinion as reflected by the high IMDB score is wholly justified.
Giorgio Ganis • March 8, 2019 11:33 AM
Probably anyone with Tourette’s would be flagged as a shoplifter. Funny that similar body signs were used by failed programs trying to detect “malintent” at airports.
albert • March 8, 2019 11:42 AM
This is to be expected. Ai is the New Fad. It’s like the fad to computerize everything. Or the fad to Internet-connect everything.
False positives will be a field day for the lawyers.
I used to work in retail sales. To bring the security teams running, all you had to do was stare at the cameras. But the security guys in those days were smart and polite, not like the minimum-wage rent-a-cops of today, who act like they’re taking down terrorists.
Nowadays, it seems like everyone has deep-rooted psychological problems.
Oh well.
Two effective anti-shoplifting* methods I’ve seen in high-crime areas: 1) A guy with a shot gun, sitting in a parapet above the floor. 2) All merchandise and clerks locked behind bulletproof glass.
*primary function was to discourage robberies.
. .. . .. — ….
vas pup • March 8, 2019 1:49 PM
@all: behavioral detectors are used for airport security and in similar environment as well as alert/warning tool for human folks responsible for security.
Israel is developing similar detectors in particular.
I guess it is good for any environment where violent conflict could evolved out of peaceful protest/gathering.
When human life is at stake, behavioral detectors are more useful versus shoplifting, but everything is finalized by cost-benefit analysis in each particular case.
As usual, ‘devil’ is not primary in technology itself, but its brainless application/utilization.
albert • March 8, 2019 2:34 PM
@vas pup,
“…As usual, ‘devil’ is not primary in technology itself, but its brainless application/utilization….”
Agreed!
I might add: never underestimate the skill of a effective sales team. What could be easier than selling security products to businesses with limited security experience?r
. .. . .. — ….
Motherfuckaaaa • March 10, 2019 3:37 AM
Another example of a deeply anti-social invention. As long as capitalism persists, artificially created resource constraints and restrictions of the basic necessities needed to survive are denied to a whole lot of individuals due to how the ponzi-scheme currently in place works, which quite naturally leads to the development of more alternative means of acquisition.
Elon • March 10, 2019 9:59 AM
Criminologists say that statistically, workers steal more than customers in retail environments. So is this technology for ignorant managers who don’t know the stats?
At my local Trader Joes, I overheard worker A saying he’s going in front of a certain judge again. Coworker B said, Oh yeah, tell him I said Hi. True story.
Anyone else remember the TSA’s widely-publicized cameras that were going to monitor airline passengers and then alert the flight crew if they decided any were terrorists? It was going to magically determine this from their expressions or something before they stood up with a bomb and a box cutter…
Eventually the TSA made a quiet announcement that the system was canceled.
Subscribe to comments on this entry
Sidebar photo of Bruce Schneier by Joe MacInnis.
Humdee • March 7, 2019 3:08 PM
It will only be the future in situations where there is only a minor penality for false positives…like at walmart for instance. But with smaller operators the hit to community goodwill will make it unacceptable.