The tech world is very excited, but also frightened, about AI & machine learning these days; we worry about AI/machine learning algorithms replacing doctors, lawyers, teachers, taxi drivers. Perhaps one of the most straightforward applications of AI & machine learning today would be a computer that "learns" how to control the emissions of a vehicle engine so that it can pass the EPA emissions tests. Consider the following conceptual model: a computer with a bunch (hundreds?) of sensors and a bunch of actuators (tens?) that watches over a diesel engine while it is being driven through a standard EPA emissions test. The computer can sense perhaps air temperature, humidity, engine speed, engine load, engine temperature, etc., and can control perhaps the air flow, the fuel flow, the flow of Adblue (aka DEF/ISO 22241), etc. Sensors don't cost very much, so there may also be sensors for the engine hood/bonnet being open, the position of the steering wheel, etc. We now put this system through hundreds of thousand of miles of "learning" (millions of miles if the testing & learning can be virtualized & run in parallel), so that the AI/machine learning algorithm learns to optimize inputs like fuel and Adblue while still meeting EPA testing limits. I can guarantee you that this AI/machine learning algorithm will quickly notice that the best way to optimize for the EPA test is to "cheat"—i.e., to notice that when the hood/bonnet is open and the steering wheel is straight ahead, this would be a good time to optimize NOx and other emissions, while during other conditions—hood/bonnet closed and steering wheel twisting back & forth (perhaps a curving country lane)—emissions aren't so important relative to performance. (Perhaps someone—Google might be in the best position with their work on autonomous robots and its expertise in "machine learning" for their self-driving cars—is already working on such AI/machine learning experiments for engine optimization; I'd be interested in hearing about them if anyone can send me links.) So is this AI/machine learning program "unethical" wrt the EPA tests ? Should it be fined or go to jail ? This is no longer idle speculation, as these AI/machine learning programs are "recognizing" speech, gaits, faces, writing styles, etc. Are they also "cheating" ? As automobiles become more complex, and as machine learning algorithms become more sophisticated, engine optimization computers may no longer be "programmed" by humans using coding techniques, but will be "taught" by following a long sequence of example situations and "learning" the correct responses. The DMCA may no longer be relevant to such computers, because *there is no source code* to look at, and indeed, the *binary code* may itself simply be a huge pile of random-looking floating point numbers in a *neural network*. The *only* way to check such a system will be through exhaustive (!) behavioral testing, as there won't be any source code to logically check for "cheats" and "defeats". I'm not trying to excuse the VW management that has already admitted to "cheating" on the EPA tests, but as a computer scientist, I'm not so sure where we go forward from here. We have terrific new opportunities with electric and self-driving cars, so "optimizing" the government regulation of diesel engines may simply be re-arranging the deck chairs on the Titanic.
Bloomberg Business Week, 28 Sep-4 Oct issue: Germany has a law, about state ownership of corporations, called the Volkswagen Law. Because of it, Lower Saxony owns 20% of VW, giving its Prime Minister virtual veto power over VW. Thanks to the VW law, which the EU has been fighting, many German companies have two boards of directors. There is the management board, with Executives, which answers to the Supervisory Board, run by the state, labor leaders, and shareholders. As yet, there is no evidence which part of VW leadership structure had anything to do with the emissions deception. The article does not address the car hacking cover-up.
I retired early this year. Previously I managed an IBM midrange system at my day job. The IBM OS tracked all sorts of updates, changes into a log, which we could examine for suspicious activity, and I regularly checked it for potential breaches, and some types of human error. We had some control over how much activity to log, and for how many days, because of disk space constraints, but some things, such as changes to the OS itself, we could not change the # days below a certain point. I had occasion to alter the system date, so I learned it was theoretically possible to erase the security log by advancing the date, so that all contents were now past the erase date. It was impossible to erase the fact that someone had messed with the system date. For an OS log to be meaningful, some humans need to be able to dig out what's important from the huge volume of frequently geeky entries. All the passwords were changeable, but IBM knew how to bust their own security. We found this out on the occasion of a relocation, where IBM techs could not reassemble everything correctly, could not get into some diagnostics, because of the security, which was on the hard drive which was not reconnecting properly. They needed authorization from our CEO to bust the security. Fortunately the CEO was watching the operation, so there was no delay getting this. If this is true across vendors, then there is a potential risk from former and present employees of the computer vendors. In my IBM Midrange world, there's a software monitoring package marketed under the name "Needle in Haystack" which sends alerts to IT people, when stuff happens which can have adverse impact on the enterprise. If the IT people do not respond, in a reasonable time frame, the alerts go up the management ladder. I imagine that other platforms, outside the IBM world, ought to have similar standards. Capture activity of a potentially suspicious nature, make it available to relevant people in an intelligible and timely manner. The next question is who this data belongs to, who may access it, update it - employees, regulators, vendors of the hardware & software. I recently had something weird happen with my auto, so I was re-reading the owner's manual. I encountered a statement that data is captured about the vehicle operations, and that this data belongs to the vehicle's owner.
For those who don't have long memories, in the late 1980s, there was serious attention to processor benchmarking by running compilers on well known libraries and benchmark test programs' source code. It was claimed that certain vendor's compilers (back then, compilers were often bespoke) would recognize the library or test program, and optimize the number crunching to the point where the benchmark test became worthless. Googling the terms computer+benchmark+cheating shows that this is still going on today.
You might want to keep in mind the incident (possibly multiple) of slot machines gaffed by the gaming board inspectors themselves. Back in the 1990s, but can we be sure they do not continue? [Of course not. The gambling machines are held to a `higher' standard, but it is still not a very high one. Remember, the best is the enemy of the good, but the so-called good is nowhere near good enough. PGN]
http://arstechnica.com/security/2015/09/storing-secret-crypto-keys-in-the-amazon-cloud-new-attack-can-steal-them/ Now a separate team of researchers has constructed a new method for recovering the full private key used in a modern implementation of the widely used RSA crypto system. Like the 2009 work, the new research implements a CPU cache attack across two Amazon accounts that happen to be located on the same chip or chipset. They recently used their technique to allow one Amazon instance to recover the entire 2048-bit RSA key used by a separate instance, which they also happened to control. The newer technique works by probing the last level cache of the Intel Xeon processor chipsets used by Amazon computers.
Ars Technica's story on the revelations reported today by The Intercept that the UK's GCHQ has been tracking World Wide Web users since 2007 with an operation called "Karma Police"—"a program that tracked Web browsing habits of people around the globe in what the agency itself billed as the 'world's biggest' Internet data-mining operation, intended to eventually track 'every visible user on the Internet.'" http://yro.slashdot.org/story/15/09/25/2349201/gchq-tried-to-track-web-visits-of-every-visible-user-on-internet <https://theintercept.com/2015/09/25/gchq-radio-porn-spies-track-web-users-online-identities/> <http://arstechnica.com/security/2015/09/gchq-tried-to-track-web-visits-of-every-visible-user-on-internet/>,
MIT via NNSquad http://www.technologyreview.com/view/538866/the-social-network-illusion-that-tricks-your-mind/ Today, we get an insight into why this happens thanks to the work of Kristina Lerman and pals at the University of Southern California. These people have discovered an extraordinary illusion associated with social networks which can play tricks on the mind and explain everything from why some ideas become popular quickly to how risky or antisocial behavior can spread so easily.
http://lauren.vortex.com/archive/001126.html There's a story going around today regarding an individual who was arrested and charged with assaulting a police officer when authorities arrived over a noise complaint. But cellphone video recorded by the arrestee convinced a judge that police had assaulted him, not the other way around. What's particularly unusual in this case is that the arrestee's cellphone had "mysteriously" vanished at the police station before any video was discovered. So how was the exonerating video ultimately resurrected? Turns out it was saved up on Google servers via the phone's enabled auto backup system. So the phone's physical vanishing did not prevent the video from being saved to help prevent a serious miscarriage of justice. Lawyers and law enforcement personnel around the world are probably considering this story carefully tonight, and they're likely to realize that such automatic backup capabilities may be double-edged swords. On one hand, abusive cops can't depend on destroying evidence by making cellphones disappear or be "accidentally" crushed under a boot. Evidence favorable to the defendant might still be up on cloud servers, ready to reappear at any time. But this also means that we may likely also expect to see increasing numbers of subpoenas triggered by law enforcement, lawyers, government agencies, and other interested parties, wanting to go on fishing expeditions through suspects' cloud accounts in the hopes of finding incriminating photographic or video evidence that might have been auto-backed up without the knowledge or realization of the suspects. While few would argue that guilty suspects should go free, there is more at stake here. The fact of such fishing expeditions being possible may dissuade many persons from enabling photo/video auto backup systems in the first place -- not because they plan to commit crimes, but just based on relatively vague privacy concerns. Even if the vast majority of honest persons would have no realistic chance of being targeted by the government for such a cloud search, an emotional factor is likely to be real for many innocent persons nonetheless. And of course, if you've turned off auto backup due to such concerns, video or other data that might otherwise have been available to save the day at some point in the future, instead may not be available at all. Adding to the complexities of this calculus is the fact that most uploaded videos or photos on these advanced systems are not subject to the kind of strong end-to-end encryption that has been the focus of ongoing controversies regarding proposed "back door" access to encrypted user data by authorities. Obviously, for photos or videos to be processed in the typical manner by service providers, they will be stored in the clear—not encrypted—at various stages of the service ecosystem, at least temporarily. What this all amounts to is that we're on the cusp of a brave new world when it comes to photos and videos automatically being protected in the cloud, and sometimes being unexpectedly available as a result. The issues involved will be complicated both technically and legally, and we have only really begun to consider their ramifications, especially in relationship to escalating demands by authorities for access to user data of all kinds in many contexts. Fasten your seatbelts.
Toymaker Mattel is coming out with $75.00 Hello Barbie, a wi-fi enabled doll. The little-girl owners shall press the belt buckle, and she'll ask questions, like "Where do you live?" and answer the child's questions. I am only guessing at what questions might be asked, as the list probably not yet published where I can find it. The child's conversations with Hello Barbie will be stored on Toy Talk servers, allegedly only to help Mattel improve their speech recognition software. But how long before Mattel sells this info to advertisers, or their data gets hacked? Can Hello Barbie get software updates, so she can promote future Mattel toys, more Barbie clothing, etc.?
Any bets about how long it will take for someone to hack this cyber 'red phone' / 'red skype'?
Mr. Ross's remarks regarding ad-blocking software in Risks 28.94 ends with the two questions: "Why can I not choose to block advertisements on the Internet? What is it about the Internet that mandates its advertisements on me, something other media cannot do?" Intended to highlight that the equivalent of Internet ad-blocking is allowed with radio, television, magazines and newspapers by turning the page, hitting the mute button or changing the channel, unfortunately the real answer to Mr. Ross's questions is "money". A quick Google search shows that Internet advertising is the largest single category of advertising, outpacing even television. Online advertising can be targeted as we know with much greater specificity than any other advertising outlet. Neither advertisers nor online media therefore have any incentive (and indeed extremely strong disincentives) to allow ad blocking if they have any say-so. So why would they?
Please report problems with the web pages to the maintainer