Facial Recognition: Accuracy in Marketing
March 10, 2019
Measurements of accuracy in search and retrieval, image recognition, and tagging human behavior are variable. Results wander from evaluation to evaluation. One can break a text retrieval system by including emojis or explicit cues that a statement belongs to a specific context. What does “cool” mean?
Recognition of images is equally tricky. Toss in low contrast images with individuals wearing hats, sun glasses, and motorcycle gang style bandanas. How accurate are these systems?
The way to work around the problem is to craft test sets of content or images. The idea is that a well formed test set will provide a level playing field.
The problem is, of course, that life is not a level playing field.
I read summaries of the NIST image recognition tests, the subsequent calls for control by Microsoft (the fading monopolist), and some of the comments about facial recognition (FR) systems having difficulty with certain ethnic groups.
When I read these, I recall the image from a hotel opening, I wonder if today’s systems can ID these individuals “accurately”:
The write up “Why Chinese Companies Plug a US Test for Facial Recognition” summarizes the results of a NIST test and references other FR bake offs. The results are variable.
My view is that the Chinese systems’ performance is less about accuracy and more about sending a message; namely,
China is in the game.
The point of FR may be that accuracy is less important than reminding those interested in FR that Chinese technology has caught up and may surpass US smart software.
Will FR be 100 percent accurate? Not as long as photos like this have to be figured out:
FR is making progress. Progress is incremental. That applies to the US and other countries’ systems.
Stephen E Arnold, March 10, 2019
Juicy Target: Big Cloudy Agglomerations of Virtual and Tangible Gizmos
March 9, 2019
Last week I had a call about the vulnerability of industrial facilities. The new approach is to push certain control, monitoring, and administrative systems to the cloud. The idea is that smart milling machines, welders, and similar expensive equipment can push their data to the “cloud.” The magic in the cloud then rolls up the data, giving the manufacturing outfit a big picture view of the individual machines in multiple locations. Need a human to make sure the industrial robots are working happily? Nope. Just look at a “dashboard.” If a deity were into running a chemical plant or making automobiles, the approach is common sense.
I read “Citrix Hacked and Didn’t Know Until FBI Alert.” The FBI is capable, but each week I receive email from companies which perform autonomous, proactive monitoring to identify, predict, and prevent breaches.
The write up points out
The firm attributed the attack to an Iranian group called “IRIDIUM” and says it made off with “at least 6 terabytes of sensitive data stored in the Citrix enterprise network, including e-mail correspondence, files in network shares and other services used for project management and procurement.”
The article buries this statement deep in the report:
The breach disclosure comes just three days after Citrix updated its SD-WAN offering to help enterprises to administer user-centric policies and connect branch employees to applications in the cloud with greater security and reliability. The product is intended to simplify branch networking by converging WAN edge capabilities and defining security zones to apply different policies for different users.
What’s the implication?
Forget Go to My PC vulnerabilities. Old news. The bad actors may have the opportunity to derail certain industrial and manufacturing processes. What happens when a chemical plant gets the wrong instructions.
Remember the Port of Texas City mishap? A tragic failure. Accidental.
But Citrix style breaches combined with “we did not know” may presage intentional actions in the future.
Yep, cloudy with a chance of pain.
Stephen E Arnold, March 9, 2019
Federating Data: Easy, Hard, or Poorly Understood Until One Tries It at Scale?
March 8, 2019
I read two articles this morning.
One article explained that there’s a new way to deal with data federation. Always optimistic, I took a look at “Data-Driven Decision-Making Made Possible using a Modern Data Stack.” The revolution is to load data and then aggregate. The old way is to transform, aggregate, and model. Here’s a diagram from DAS43. A larger version is available at this link.
Hard to read. Yep, New Millennial colors. Is this a breakthrough?
I don’t know.
When I read “2 Reasons a Federated Database Isn’t Such a Slam-Dunk”, it seems that the solution outlined by DAS42 and the InfoWorld expert are not in sync.
There are two reasons. Count ‘em.
One: performance
Two: security.
Yeah, okay.
Some may suggest that there are a handful of other challenges. These range from deciding how to index audio, video, and images to figuring out what to do with different languages in the content to determining what data are “good” for the task at hand and what data are less “useful.” Date, time, and geocodes metadata are needed, but that introduces the not so easy to solve indexing problem.
So where are we with the “federation thing”?
Exactly the same place we were years ago…start ups and experts notwithstanding. But then one has to wrangle a lot of data. That’s cost, gentle reader. Big money.
Stephen E Arnold, March 8, 2019
Cellebrite Products on eBay?
March 8, 2019
Quite an assertion appeared in “Cellebrite’s phone Hacking Tools Going Cheaply on eBay, Many Still Contain Unwiped Data.”
Cellebrite, a unit of Japan’s Sun Corporation, offers specialized services and hardware to law enforcement agencies worldwide. Like other cyber security and policeware vendors, Cellebrite is a secretive company.
Prices much lower than Cellebrite’s. Cellebrite, it appears, may have need to revisit its product resale guidelines for its customers. More information about Cellebrite can be found on the company’s Web site.
Stephen E Arnold, March 8, 2019
Simple Ways Intelligence is Fighting Cyber Crimes
March 8, 2019
Our world has never been more technologically advanced, that’s a fact. That also means that the digital threats have never been more dire, right? Yes and no, according to one source, who says that the technology might change but humans never do. We learned more from a recent CNBC story, “Google Infosec Head Heather Adkins: Ignore Scare Stories.”
According to the story:
“Adkins said sometimes the marketplace suffers from a “proliferation of cybersecurity professionals” offering conflicting advice on passwords, antivirus software, safety practices and so on…But the best rules for individuals looking to secure their personal information are the classics, Adkins said…Keep your software up to date, and don’t re-use the same password.”
This and many other examples show that good old fashioned foresight and detective work can still help fight cybercrime, even in this world of machine learning and nanotech. As Adkins says, let’s look forward in regards to security, but also not forget our past.
However, fear, uncertainty, and doubt sell—particularly to some executives uncomfortable with today’s business environment.
Patrick Roland, March 8, 2019
Facebook Tracking Amidst Privacy Assertions
March 7, 2019
Privacy International published “Guess What? Facebook Still Tracks You on Android Apps (Even If You Don’t Have a Facebook Account).”
I am not particularly surprised. The chatter about Facebook and its privacy initiative is one of those “pivot” plays. Talk is cheap, unlike online advertising.
The write up states:
seven apps, including Yelp, the language-learning app Duolingo and the job search app Indeed, as well as the King James Bible app and two Muslim prayer apps, Qibla Connect and Muslim Pro, still send your personal data to Facebook before you can decide whether you want to consent or not. Keep in mind: these are apps with millions of installs.
There are some recommendations in the write up. DarkCyber suggests you read these before spending much time on statements like this one from Facebook: “A Privacy Focused Vision for Social Networking.”
Stephen E Arnold, March 7, 2019
New Field Manual about Social Media Deception
March 7, 2019
How effective are false personas and social media? The answer is, “Remarkably effective.” A new report published by the US Army provides a wealth of information about online deception. The 100 page publication is “FM 3-13.4 Army Support to Military Deception.” The document provides a review of the fundamentals of deception, information about planning a deception operation, and data about preparation and execution. An example of the information contained in the report is a useful review of the terminology of deception. DarkCyber requires these types of glossaries essential to researching additional information.
Without the key terms or jargon, queries about software, methods, and activities are difficult to formulate. Equally valuable is the checklist of “principles.” The information provides a series of reminders about specific operational considerations required for a successful deception campaign. The list of techniques is one of the first summaries of the spectrum of functions associated with deception operations.
The report captures a post operation evaluation checklist. One weakness is the report does not operational examples focused on social media like Facebook, Instagram, Twitter, and LinkedIn. With NATO’s social media phishing of its own soldiers proving effective, more attention is warranted on what can be accomplished at this time with widely used online services. DarkCyber recommends this document because deception activities will become an increasing important tool in law enforcement and intelligence activities.
Stephen E Arnold, March 7, 2019
AI Develops its Own Language and Problems
March 7, 2019
Opportunity or new challenge: A language created by smart software without a human’s knowledge, input, or understand has arrived.
The rapid pace of artificial intelligence development is not news. For years we have seen these rapid advances on the horizon, but only recently have we reached that tipping point. One of the most interesting recent areas of growth were language, which can become an issue for some professionals. We discovered more from the recent Psychology Today story, “Google and Facebook AI Make New Linguistics Discovery.”
According to the story:
“The researchers found that “the success rates between self-play and paired-play are indistinguishable from each other, strongly implying that a common, shared language emerges as a social convention if and only if we have more than two language users,” and all “that is needed in order for a common language to emerge is a minimum number of agents.”
Machine generated “language”, known only to the machine, can be used to generate messages which may pose additional challenges to code breaking tools and human specialists.
Patrick Roland, March 7, 2019
Factualities for March 6, 2019
March 6, 2019
Believe the data or not. We believe everything we read on the Internet, including the ads, Facebook PR, and statements from any online source that presents information. That’s the way it should be.
1,600. The amount of horsepower in the Koenigsegg Jesko. Source: CNet
1,600. Number of Teslas held in Chinese customs. Sources: Caixing Global
28 days. The average life of a scooter in Louisville, Kentucky. The interval is almost as long as Google’s commitment to high speed Internet in the city. Source: Oversharing
36. The number of months Moldovan developers reported flows of misinformation and fake news to Facebook. Facebook took action in February 2019. Source: Buzzfeed
23 million. Number of times Chinese nationals with a low social credit (social behavior) score were stopped from buying train or airplane tickets. Source: AP
40 percent. Number of artificial intelligence start ups in Europe which are not involved in AI. Source: The Inquirer
45 percent. The percentage of Americans who have had a brother or sister, parent, spouse or child spend time in jail or prison. Source: Cornell University
32 billion. Number of “compromised credentials” in the SpyCloud database. Source: Venture Beat
18 percent. Percentage of minors in Facebook’s for-pay monitoring service. Source: TechCrunch
$23 million. Incentives to be paid to Amazon has to take “a certain amount” of office space in the once thriving Crystal City complex. Source: Washington Post
43. The number of patents Apple was granted on January 1, 2019. Source: Patently Apple
70,000. Number of Chromecasts seized by hackers to promote PewDiePie. Source: Android Authority
Stephen E Arnold, March 6, 2019
IBM and Its Farm Team Model
March 6, 2019
IBM and SUNY Poly have pooled their talents to create an innovating institute: a brand new artificial intelligence lab in Albany, New York. Times Union has the details in the article, “IBM, SUNY Poly Creating An Artificial Intelligence Center In Albany.” The new AI lab is only one part of a $2 billion commitment IBM has made to New York University.
The deal outlines that Empire State Development will provide SUNY with a $300 million grant over five years for the AI Hardware Center. As a trade, IBM will remain at SUNY Poly’s Center for Semiconductor Research until 2023. IBM’s spending over the period will amount to $2 billion and will make New York one of the top places for AI.
We learned:
“ ‘New York has always been at the forefront of emerging industries, and this private sector investment to create a hub for artificial intelligence research will attract world-class minds and drive economic growth in the region,’ Gov. Andrew M. Cuomo said in a statement. “Artificial intelligence has the potential to transform how we live and how businesses operate, and this partnership with IBM will help New York stay on the cutting edge developing innovative technologies.’”
The AI Hardware Center is only one of many AI institutes, among them are Rensselaer Polytechnic Institute, Applied Materials, and Tokyo Electron. IBM will also spend $30 million on AI programs at SUNY schools, while SUNY will contribute $25 million as well. SUNY wants all of its schools to have closer collaboration and it will be part of a new education model.
Is this a “farm to table” approach to innovation? Watson, what’s the answer?
Whitney Grace, March 6, 2019
-
- Subscribe to Beyond Search
Feature archive
News archive
-
