Association for Computing Machinery
Welcome to the April 18, 2016 edition of ACM TechNews, providing timely information for IT professionals three times a week.

Updated versions of the ACM TechNews mobile apps are available for Android phones and tablets (click here) and for iPhones (click here) and iPads (click here).

HEADLINES AT A GLANCE


Grand Challenges for Cultural Heritage Databases, From Preservation to Best Practices
Yale News (04/14/16) Jim Shelton

The digital revolution that spurs social movements and new technology also is changing the preservation of cultural heritage. Computer scientists and engineers are working with governments, museums, law enforcement agencies, and private groups to collect data on a wide range of fields. The effort to collect the data, which is being used to disseminate cultural heritage information on a global scale, last week was highlighted at a panel discussion for the eighth United Nations Global Colloquium of University Presidents (UNGC), hosted this year by Yale University. "Technologists and computer scientists like myself build tools to empower people and amplify human effort," says Yale professor Holly Rushmeier, who moderated the discussion. Cultural heritage data includes images, content, and audio from cellphones, social media, and satellites. Users around the world can engage with the information, but there is little standardization in the way such content is presented and authenticated. The panelists raised several issues relating to cultural heritage and technology. For example, Getty Conservation Institute director Tim Whalen presented an open source Web platform that can preserve data about cultural monuments, buildings, art, and historic sites. The panel also discussed the impact of digitization and three-dimensional printing on the way cultural heritage is perceived.


The Autonomous Car's Reality Check
The Economist (04/16/16)

The German mapping firm HERE is sponsoring trips in 32 countries by vehicles equipped with instruments to plot out driving routes in three dimensions (3D) to create high-definition (HD) maps as aids to autonomous cars. HERE's vehicles, which are driven by people, record latitude, longitude, and elevation readings along with yaw, pitch, and roll, with laser-scanning of distances and multiple cameras capturing panoramic images. The accumulation of at least 100 GB of data for each trip is fed into the HD map model. The maps are seen as a way for driverless cars to orient themselves precisely while overcoming the shortcomings of other technologies, such as laser-surveying sensor systems and commercial global-positioning systems. Such tools' precision can be constrained by poor visibility or limited accuracy or coverage, whereas HD maps can position a car within centimeters via a localization layer and sensors working together. HERE also is experimenting with a system that employs artificial intelligence (AI) to recognize features from cameras and LIDARs. A key challenge is analyzing and processing the data from numerous vehicles, and HERE developed machine-learning algorithms to automatically identify lane markings and pavement edges, while AI systems recognize other features from still photos and people refine the results and check for errors.


Half of MOOC Users in Developing Countries Get Certificates
Campus Technology (04/12/16) Michael Hart

A survey of 1,400 massive open online course (MOOC) users and 2,250 non-users from ages 18 to 35 in Colombia, the Philippines, and South Africa found the use of the technology is much different in emerging countries than in the U.S. In some ways, the use of MOOCs in emerging countries is more in line with what the original proponents had in mind, according to researchers at the University of Washington. Nearly half of those who had taken a MOOC had received certification in at least one course, and the rate was higher--70 percent--for those who were employed. Compared to the U.S., more MOOC users in the developing countries indicated they did not have a college education, suggesting they were using the courses to advance their careers or education. "Many people assumed that in developing countries, MOOCs would only be used by the rich and well-educated," says Maria Garrido with the University of Washington's Technology and Social Change Group. The top reasons for taking MOOCs were to gain skills to perform their jobs better, prepare for additional education, and obtain professional certification. In more economically advanced countries, 66 percent use MOOCs for personal fulfillment, according to a recent study of Coursera users.


Two-Factor Authentication Bypassed in Simple Attacks
Security Week (04/12/16)

VU University Amsterdam researchers have demonstrated practical attacks against both Android and iOS devices, showing how a man-in-the-browser (MitB) attack can be elevated to bypass two-factor authentication (2FA) mechanisms. The increased usage of smartphones and people's tendency to keep applications synchronized across multiple devices makes phone-based 2FA useless, according to the researchers. The synchronization of apps means once an attacker can access a user's computer, the smartphone can be compromised to bypass the security mechanism. Since 2FA relies on the idea of segmentation to protect against attacks and malware, the process of integrating apps among multiple platforms is negating its benefits and exposing users. In addition, because of synchronization, once a victim's computer has been breached, the attacker can engage in MitB attacks and perform illegal operations 2FA should have prevented. "By exploiting certain 2FA synchronization vulnerabilities, we show that mobile phone 2FA as used by many online services for secure authentication, including financial institutions, can be easily bypassed," the researchers warn. They demonstrated the attack by leveraging Google Play's remote app installation feature, which enabled them to install a specifically designed vulnerable app on the victim's Android devices. The iOS attack was designed around a new OS X feature that enables users to synchronize their short-messaging-services messages between the iPhone and Mac computer.


Stanford and Wikimedia Researchers Create a Tool to Boost Article Creation in Local Language Wikipedias
Stanford Report (04/14/16) Tom Abate

Researchers at Stanford University and the Wikimedia Foundation have developed a tool designed to help editors in different linguistic communities identify gaps by identifying the most important articles that are still unavailable in a given language. Multilingual editors can use these recommendations to locate an article in a second language they are familiar with and get additional assistance in order to translate the article for local Wikipedia readers. The team first created lists of every article in each language, and then cross-referenced them to determine which articles were not available in which languages. The researchers then estimated the importance of each missing article according to cultural and geographic relevance. They used the 4.9 million articles in English Wikipedia to find those that were missing relative to the 1.6 million articles in French Wikipedia. The team then selected the 300,000 most important English articles missing from the French archive, which were randomly divided into three groups of 100,000 articles each and circulated to selected editors. The researchers found directing one group of 6,000 editors toward five random missing articles doubled the organic article creation rate. This rate tripled for a second group where the five suggestions were tailored to the editors' interests.


Location Data on Two Apps Enough to Identify Someone, Says Study
Columbia University (04/13/16) Kim Martineau

Individuals can be identified with a high degree of confidence by matching their movements across two datasets, according to a team of researchers at Columbia University and Google. The team demonstrated that geotagged posts on just two social media apps are enough to link accounts held by the same person. The researchers used an algorithm that calculates the probability one person posting at a given time and place also could be posting in a second app at another time and place. The research, which was presented at the World Wide Web (WWW 2016) conference in Montreal, raises privacy concerns. "What this really shows is that simply removing identifying information from large-scale datasets is not sufficient," says Yves-Alexandre de Montjoye, a research scientist at the Massachusetts Institute of Technology Media Lab who was not involved in the study. Columbia undergraduate students also built a related tool, You Are Where You Go, to enable individuals to audit their social media trail. The tool retraces users' steps and makes relatively accurate inferences about their age, income, and whether they have children.


Clinician-Mimicking Program Could Improve Brain Injury Analysis
Imperial College London (04/12/16) Colin Smith

Imperial College London (ICL) researchers have developed Deep Medic, software that mimics how a clinician analyzes brain scans to identify lesions, which may indicate traumatic brain injury (TBI). The researchers see Deep Medic as a tool to help clinical researchers understand more about TBI. Deep Medic is an artificial neural network, a predictive algorithm the researchers trained to look for abnormalities in brain tissue. Deep Medic develops a three-dimensional image of a brain to pinpoint where lesions are located, and this year the researchers will use Deep Medic in a large-scale European study called CENTER TBI. The study will collect data from 5,000 patients at 30 hospitals across Europe, and Deep Medic will identify TBI in the patients' brain scans. The study could reveal new insights into TBI, and lead to the development of more effective and efficient therapies for patients at lower costs. The researchers have primarily used Deep Medic to analyze magnetic resonance imaging scans associated with TBI, but the technology also could be used to detect brain tumors in patients with cancer and lesions caused by strokes. "What is exciting about Deep Medic is that it can analyze a scan in a matter of minutes, which has the potential to improve research outcomes, and which may ultimately have benefits for patients with brain injuries," says ICL researcher Ben Glocker.


Robots Can Help Young Patients Engage in Rehab
The Conversation (04/10/16) Chris McCarthy; Adam Scheinberg; Felip Marti; et al.

Researchers from the Swinburne University of Technology are exploring the use of social robots as therapeutic aids in pediatric rehabilitation. Over the past 12 months, the highly engaging humanoid robot Nao has worked with more than 30 individual patients at the Royal Children's Hospital in Melbourne. The 23-inch-tall robot features life-like gestures and is programmed to perform the roles of motivator, demonstrator, and instructor. Equipped with a range of sensors, Nao attentively tracks movement, provides feedback on technique, and reacts to specific milestones with gestures and words of encouragement. Physiotherapists have generally noted improvements in both patient compliance and mood when Nao is present compared to sessions when it is not. Nao's persona as a child-like peer instead of instructing adult and its novelty factor have contributed to its success. The researchers say the rapport with patients should improve as more advanced machine-learning, natural-language processing, and computer-vision algorithms build up Nao's situational awareness.


Commonly Used Strategy for Website Protection Is Not Waterproof
KU Leuven (04/12/16)

Cloud-based security providers often protect themselves against cyberattacks by diverting incoming Web traffic via their own infrastructure, but the success of this strategy depends on how well the website's original Internet Protocol (IP) address can be shielded. However, if that IP address can be retrieved, KU Leuven researchers say protection mechanisms can be easily bypassed. To protect against this flaw, the researchers set up the first large-scale research effort focused on exploring vulnerabilities in the Domain Name System (DNS) redirection strategy. The researchers developed CLOUDPIERCER, a tool that automatically tries to retrieve websites' original IP addresses based on eight different methods, including the use of unprotected subdomains. Nearly 18,000 websites, protected by five different providers, were subjected to the researchers' DNS redirection vulnerability tests. "The results were pretty confronting: in more than 70 percent of the cases, CLOUDPIERCER was able to effectively retrieve the website's original IP address, thereby providing the exact info that is needed to launch a successful cyberattack," says KU Leuven researcher Thomas Vissers. "This clearly shows that the DNS redirection strategy still has some serious shortcomings." The researchers shared their results with the cloud-based security providers involved in the study, but they note people can use CLOUDPIERCER to test their own websites.


Humanoid Robotics and Computer Avatars Could Help Treat Social Disorders
University of Bristol News (04/11/16)

Humanoid robotics and computer avatars could help rehabilitate people suffering from social disorders, according to a collaborative research team from the universities of Bristol, Exeter, Montpellier, and Naples Federico II. The researchers say their breakthrough is based on the theory of similarity, which suggests it is easier to interact socially with someone who looks, behaves, or moves like people do. The research uses the principles of dynamical systems and feedback control theory to embed the avatar with enough intelligence to sync and respond to the motion of the patient. The system enables an avatar to interact with a patient while playing a version of the mirror game, in which two players try to copy each other's motion while moving colored balls horizontally on a string. The avatar initially is like an alter ego, created to look and move like the patient to enhance their feelings of attachment, but over time it changes to become less similar, which helps social rehabilitation. The results indicate players sharing similar movement features, or motor signatures, interact and coordinate better. "It is very challenging to build an avatar that is intelligent enough to synchronize its motion with a human player, but our initial results are very exciting," says University of Bristol professor Mario di Bernardo.


UA Team Revs Up Transportation's Internet
UA News (AZ) (04/08/16) Jill Goetz

University of Arizona (UA) researchers say they are developing vehicle-to-infrastructure and vehicle-to-vehicle communications systems that could significantly change how traffic is managed and how drivers experience the road. The researchers have equipped a 2.3-mile stretch of road in Anthem, AZ, with dedicated short-range communications radios (DSRCs) mounted on demo buses and emergency vehicles, atop light poles, and in roadside equipment boxes at intersections. The wireless devices work with traffic signal controllers and optimization and scheduling algorithms to manage traffic signals in a new way. The signals used algorithms developed by the researchers to prioritize approaching passenger, transit, emergency, and commercial vehicles. The UA researchers worked with colleagues from the California Partners for Advanced Transportation Technology program to enable communication between traffic lights and other vehicles, and made it possible for all of the vehicles to communicate with each other in real time. "This capability for multiple vehicles to simultaneously request and receive traffic signal priority is what makes our system unique," says UA professor Larry Head. The researchers currently are pursuing funding for the project's next phase, which will include installing DSRCs in up to 20 percent of vehicles used by Anthem residents.


Microsoft Research Chief: AI Is Still Too Stupid to Wipe Us Out (and Will Be for Decades)
TechRepublic (04/15/16) Nick Heath

Microsoft Research Cambridge laboratory director Chris Bishop dismisses the fear artificial intelligence (AI) is on the cusp of overtaking human intelligence, and says it will continue to lag human performance for decades to come. "Yes, deep learning has achieved human-level performance in object recognition, but what does that mean?" Bishop asks. "It means the machine makes about the same number of errors as the human." Bishop stresses even vaunted examples of machine intelligence, such as Google DeepMind's Go-playing system, have to be understood within the context of the immense time and manpower invested in their development. He also cites the common misconception that machine-learning systems' ability to perform some of the individual tasks people can do means they are on the brink of matching more general human abilities. Imperial College London professor Maja Pantic says this myth is debunked by the fact that building generalized systems capable of solving any possible problem was proved impossible. Although Bishop downplays fears of homicidal AIs annihilating humanity, he acknowledges more mundane dangers and risks fundamental to the technology. For example, Bishop notes the opaque nature of deep neural networks raises the possibility the AI's decisions could be shaped by unknown biases stemming from the vast amount of data on which such systems are trained.


Abstract News © Copyright 2016 INFORMATION, INC.
Powered by Information, Inc.


To submit feedback about ACM TechNews, contact: [email protected]
Current ACM Members: Unsubscribe/Change your email subscription by logging in at myACM.
Non-Members: Unsubscribe