Welcome to the March 15, 2017 edition of ACM TechNews, providing timely information for IT professionals three times a week.

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

Row of computer servers NSA, DoE, See China Nearing Supercomputing Leadership
Patrick Thibodeau
March 15, 2017

High-performance computing (HPC) experts at the U.S. National Security Agency and the Department of Energy warn of the high probability that China will assert global domination of the supercomputing field as soon as 2020 without an aggressive increase in U.S. investment. Their report speculates China's ascension not only imperils national security, but also U.S. leadership in high-tech manufacturing. The report notes China's 93-petaflop Sunway TaihuLight system is comprised of "homegrown" elements, and is used for cutting-edge research. Hyperion Research's Steve Conway says the report is a wake-up call for government officials about "the dangers of taking U.S. HPC leadership for granted when other nations, particularly China, are intent on seizing global leadership of the market for supercomputers." The report also notes China's HPC investments are designed to improve the country, raise its population out of poverty, and establish a world-leading economic power.

Full Article

Illustration of brain with microchip DeepMind Finds Way to Overcome AI's Forgetfulness Problem
Jeremy Kahn
March 14, 2017

Researchers at Google's DeepMind unit in the U.K. say they have addressed neural-network systems' inability to remember because they overwrite previously learned knowledge whenever they receive new data. They say their breakthrough should improve software's ability to learn a sequence of tasks and make artificial intelligence systems more versatile. "We have shown it is possible to train a neural network sequentially, which was previously thought to be a fundamental limitation," says DeepMind's James Kirkpatrick. The researchers' Elastic Weight Consolidation algorithm computes how important each neural-network link is to the task it has just learned, and assigns each link a mathematical weight proportional to its importance. The weight slows the rate at which the value of that specific network node can be changed, thus retaining knowledge while learning a new task. DeepMind's research also could have significant ramifications for the study of synapses in the brain.

Full Article

Illustration of sound waves It's Possible to Hack a Phone With Sound Waves, Researchers Show
The New York Times
John Markoff
March 14, 2017

Researchers at the University of Michigan (U-M) and the University of South Carolina on Tuesday demonstrated smartphone hacking via sound waves. They added fake steps to a Fitbit fitness monitor and played a "malicious" music file from the speaker of a smartphone to commandeer its accelerometer. U-M professor Kevin Fu says this exploit let them manipulate software that relies on the smartphone. In testing 20 accelerometer models from five manufacturers, the hack method affected the information or output from 75 percent of the devices tested and controlled the output in 65 percent of the devices. Fu also has investigated the cybersecurity risks of medical devices, including a demonstration of the potential to wirelessly induce deadly heart rhythms into a pacemaker. The researchers say their study offers insight into the innate cybersecurity challenges of complex systems in which analog and digital elements can interact unpredictably.

Full Article
New Japanese Supercomputing Project Targets Exascale
HPC Wire
Tiffany Trader
March 14, 2017

A new supercomputing project in Japan is targeting exascale capacity. The project's leader, ExaScaler CEO Motoaki Saito, has founded three high-performance computing firms that will each tackle a key component in the development of an exascale system, including a manycore processor, liquid cooling, and a three-dimensional (3D) multilayer memory system. Keio University professor Tadahiro Kuroda will contribute a high-capacity, low-power 3D integrated circuit he developed, while one of Motoaki's companies will provide liquid carbon fluoride cooling technology. With this approach, the build-out will consist of 18 connected supercomputers supporting a 24-petaflops system to be installed at the Japan Agency for Marine-Earth Science and Technology's Yokohama Institute for Earth Sciences. Japan News says the development team aims to realize the fastest computing speed by June, "which would make the computer the third-fastest in the world." Junichiro Makino at the RIKEN Advanced Institute for Computational Science, recipient of the ACM Gordon Bell Prize for 2012, says this milestone could have "revolutionary" implications for next-generation supercomputers.

Full Article
Doubts About Whether Internet Filters Protect Teenagers Online
University of Oxford
March 14, 2017

Researchers at the Oxford Internet Institute at Oxford University in the U.K. analyzed data from the U.K. Office of Communications (Ofcom) based on 1,030 interviews in the homes of 515 teenagers and their parents about whether they used technical tools to control or manage access to online content. The researchers found about 14 percent of the teenagers interviewed reported they had at least one significant negative experience online in the past year, while 8 percent said they had been contacted by someone online who they did not know and wanted to be their friend. In addition, the survey found 66 percent of parents said they did not use Internet filters, indicating the use of Internet filtering did not appear to mitigate the risk of young people having unpleasant online experiences and the technical ability to circumvent these filters had no effect on the likelihood of such experiences.

Full Article
Nanoscale Logic Machines Go Beyond Binary Computing
Lisa Zyga
March 14, 2017

An international team of academic researchers has built minuscule logic machines that physically model computational problems and exploit the innate randomness of nanoscale systems, instead of relying on binary switching. The machines are formed from phosphorus atoms embedded in a silicon crystal, and each atom can occupy one of four possible states between which it is constantly shifting, based on a specific set of probabilities. These likelihoods correspond to random electron movements caused by quantum tunneling. "Our approach shows the possibility of a new class of tiny analog computers that can solve computationally difficult problems by simple statistical algorithms running in nanoscale solid-state physical devices," says Francoise Remacle at the University of Liege in Belgium. He notes the biggest challenge is realizing complete control of the position of dopant atoms in the crystal with atomic precision and the design of their transport properties.

Full Article
Naming Computers Endangers Privacy, Say 'Net Standards Boffins
The Register (UK)
Richard Chirgwin
March 14, 2017

A group of advocates for Internet standards contend an overabundance of online protocols disclose enough information to make hostnames a privacy risk. "It is common practice to use the hostname without further qualification in a variety of applications from file sharing to network management," they note. "Hostnames are typically published as part of domain names and can be obtained through a variety of name lookup and discovery protocols." In a paper, the team says experiments at an Internet Engineering Task Force meeting demonstrated that with sufficient hostnames in a database and access to other datasets, "the identification of the device owner can become trivial given only partial identifiers in a hostname." Protocols they identified as leaking hostnames include the dynamic host configuration protocol, various aspects of the domain name system, link-local multicast name resolution, and NetBIOS over TCP. They suggest applying MAC address randomization principles to hostnames.

Full Article

A tsunami wave over a city Scientists Develop a System That Predicts the Behavior of Tsunamis in Less than 10 Minutes
University of Granada (Spain)
March 14, 2017

Researchers at the University of Granada in Spain say they have developed a simulator that needs only 10 minutes to predict the behavior of tsunamis generated by landslides. The system reduces the time spent in calculating different situations up to 60 percent, making it possible to immediately obtain information and facilitate more effective response from authorities and rescue teams. The model accurately predicts the effects of the wave and performs a simulation before the events occur in real life. The simulation provides information on the time the tsunami will take to land, the magnitude and height of the wave, the coastal penetration, and the flood that it would cause. The researchers focused on an event from 1957 in Lituya Bay, Alaska, when 30 million cubic meters of glacier fell into the water and generated a wave that flooded zones located more than 500 meters above sea level.

Full Article
Is Reliable Artificial Intelligence Possible?
Swiss Federal Institute of Technology in Lausanne
Hillary Sanctuary
March 14, 2017

Artificial intelligence (AI) technology should be openly available, according to researchers at the Swiss Federal Institute of Technology in Lausanne (EPFL) in France. "We have to hold AI accountable, and the only way to do this is to verify it for biases and make sure there is no deliberate misinformation," which is not possible if the AI is privatized, according to EPFL researcher Marcel Salathe. Last year, EPFL researchers created an algorithm to recognize plant diseases, and Stanford University researchers recently showed AI can be trained to recognize skin cancer. Although these diagnostic tools use datasets of images to train and learn, the datasets can be influenced to prevent deep-learning algorithms from classifying images correctly. In addition, deep neural networks are highly vulnerable to visual perturbations that cause the AI to misclassify images. These vulnerabilities highlight the importance of certifying AI technology and monitoring its reliability.

Full Article
New Multi-Device System for Handling Emergencies With Information From Social Networks
Carlos III University of Madrid (Spain)
March 13, 2017

Researchers at the Carlos III University of Madrid (UC3M) in Spain have unveiled a prototype multi-device system that enables citizens to participate in emergency management using information from social networks. The platform also gathers, filters, and aggregates data from Twitter and other social networks to improve emergency responses. "Our system addresses the relevance of the information via an innovative method based on ontologies, which enables us to filter 'tweets' and extract significant subjects according to their semantics," says UC3M's Teresa Onorati. The system filters the most important information and represents it via different visualizations so emergency operators can make faster and easier decisions. The tool can be incorporated within an emergency center, and UC3M professor Paloma Diaz says it "prevents distracting operators and wasting resources and time, given that the visualizations represent how the situation is evolving."

Full Article
Total Protection for Online Pictures and Videos
Ben-Gurion University of the Negev (Israel)
Andrew Lavin
March 13, 2017

Ofer Hadar, a professor at Ben-Gurion University of the Negev in Israel, has developed a new method, called the Coucou Project, to provide virtually flawless protection against cyberattacks launched via online videos or images. He developed algorithms that can prevent attackers from infiltrating and extracting information through videos or pictures. "The idea is to manipulate the file's 'payload' to remove the malicious code without damaging the data quality," Hadar says. His strategy addresses several scenarios; for example, once the user uploads an image or a video to a social network, the malware embeds the classified information into the uploaded content. In another, the hacker uploads infected content to a social network or any other shared server where the malware can extract and execute the malware. "Preliminary experimental results show that a method based on a combination of Coucou Project techniques results in virtually 100-percent protection against cyberattacks," Hadar says.

Full Article
HKBU Scholar Invents World's First 'Lip Password:' A Patented Double Security System for Identity Authentication
HKBU eNews
March 6, 2017

Researchers at Hong Kong Baptist University have developed "lip motion password" (lip password), new technology that utilizes a person's lip motions to create a password. The system verifies a user's identity by simultaneously matching the password content with the underlying behavioral characteristics of lip movement. The researchers say the technique has advantages over conventional security access control methods--a lip password can be used for speaker verification because it can detect and reject a wrong password spoken by the user or the correct password spoken by an imposter, while verification based on a combination of lip motions and password content ensures that access control is twice as secure. In addition, lip movements are less susceptible to background noise and distance, and can be used by the speech-impaired. A user also can easily reset the lip password to strengthen security, and there is no language boundary.

Full Article
Why We Should Not Know Our Own Passwords
The Conversation
Megan Squire
March 9, 2017

There is active research transpiring in the area of unknowable password development, with Elon University professor Megan Squire citing several notable projects. Researchers at California State Polytechnic University in 2016 proposed a solution that measures a person's unique brain chemistry response while listening to their choice of soothing music. Squire notes this biometric response is incorporated into the user's login process. If a user is under stress, they cannot relax enough to match their previously measured "chill" state, and the login will not go through. Meanwhile, Google's Project Abacus proposes replacing the traditional password with a Trust Score, a proprietary blend of identifiable characteristics determined by Google. The score includes biometric factors such as typing patterns, walking speed, voice patterns, and facial expressions. "If the Trust Score falls below a certain threshold, say by observing a strange typing pattern or an unfamiliar location, the system will require the user to enter additional authentication credentials," Squire says.

Full Article
ACM Future of Computing Academy

Association for Computing Machinery

2 Penn Plaza, Suite 701
New York, NY 10121-0701

ACM Media Sales

If you are interested in advertising in ACM TechNews or other ACM publications, please contact ACM Media Sales or (212) 626-0686, or visit ACM Media for more information.

To submit feedback about ACM TechNews, contact: [email protected]