Association for Computing Machinery
Welcome to the September 12, 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).


Self-Driving Cars Can Learn a Lot by Playing Grand Theft Auto
Technology Review (09/12/16) Will Knight

Several research groups are using the "Grand Theft Auto" video game to educate algorithms with potential application to self-driving vehicle navigation. The researchers say the game's scenic elements are extremely realistic, and can be used to produce data on a par with that generated using real-world imagery. A team from Intel Labs and Germany's Darmstadt University created a software layer positioned between the game and a computer's hardware, automatically classifying different objects in the game's road scenes. The team says this provides the labels that can be entered into a machine-learning algorithm, enabling it to recognize cars, pedestrians, and other objects displayed either in the game or on an actual street. A key challenge in artificial intelligence that applies particularly to automated driving is the work involved in collecting and labeling real-world imagery, and the annotation required is not easily scalable at present, says University of British Columbia (UBC) postdoctoral researcher Alizera Shafaei. He and UBS professor Mark Schmidt demonstrated video games can not only be used to train a computer-vision system, but also easily vary the environmental conditions found in training data. "We showed that this synthetic data is almost as good, or sometimes even better, than using real data for training," Shafaei says.

Japan's Post-K Computer Hits 1-2 Year Speed Bump
HPC Wire (09/08/16) Tiffany Trader

Japan's Post-K supercomputer has been delayed by one to two years, and will not be deployed until 2021 or 2022, according to RIKEN AICS project leader Yutaka Ishikawa. Ishikawa says the additional time is needed to ensure sufficient processor volume. He notes the new timetable also shifts Japan's exascale horizon closer to that of the U.S., which aims to have a productive exascale computer by no later than 2023. The Japanese government has so far budgeted 110 billion yen (about $1 billion) toward the Post-K computer, with an additional 20 billion yen (about $200 million) being funded by Fujitsu. However, the company is currently negotiating to increase its compensation with a final budget expected sometime next year. Fujitsu plans to use the ARMv8-A chips in the Post-K computer, a switch from the SPARC64 chips used in the K Computer. "With the ARM architecture, we expect wider community support, leading to mature tools, compilers, and systems software," Ishikawa says. He notes the new system will serve as an essential tool for solving problems in the areas of bioscience, disaster prevention, environmental issues, energy, and manufacturing.

A Chip-Scale Source for Quantum Random Number Generators
IEEE Spectrum (09/08/16) Rachel Courtland

Spanish and Italian researchers have created an integrated circuit that can generate true random numbers by utilizing the unpredictable nature of quantum mechanics. They say the approach could be useful to engineers who hope to keep financial transactions and other communications secure. The new chip is made using standard fabrication techniques for photonic integrated circuits, then combined with a small, pulsed indium phosphide laser to infuse the system with randomness. Below a certain energy threshold, the laser emits a small number of photons via spontaneous emission, which creates light with random phase. This randomness impacts the ultimate phase of the light the laser emits when it is above that threshold, once the stimulated emission starts to dominate, says Valerio Pruneri of the Institute of Photonics Sciences in Spain. He notes this process results in laser light with a random pulse, and the researchers converted these random phases into a useful system by mixing the pulsed light with a second indium phosphide laser on the chip. Pruneri says the phase of the first laser's pulse affects how light from the two laser sources interfere with one another, creating certain brightness differences that can be analyzed by a photodetector. Pruneri also says this quantum "entropy source" can then be used to generate random numbers at about 1 Gbps.

Computers That Can Argue Will Be Satnav for the Moral Maze
New Scientist (09/07/16) Gilead Amit

Computers that formulate arguments could lead to "research engines" to inform decision-making across a wide range of fields. Researchers at the IBM Haifa Research Lab in Israel are working on a project to see if IBM's Watson supercomputer can be upgraded from a fact-checking machine to an argument generator. A team led by IBM Haifa's Noam Slonim first sought to classify an argument's logical definition, training Watson to distinguish between claims and generic statements by sifting through Wikipedia. Slonim says they have started identifying key features to establish such distinctions, while later tasks include flagging evidence supporting claims and teaching the computer to differentiate between--and assign weight to--anecdotal evidence and expert testimony. Another challenge is to train the computer to master the ability to shape arguments by appealing emotionally "to facilitate and encourage good quality argumentation and debate," says researcher Chris Reed of Britain's University of Dundee. His team is focused on seeking out solid arguments, and then deconstructing and repurposing them to train artificial intelligence (AI) to argue like a human. Reed says they are dissecting ethical debates on a BBC radio program to classify arguments and their interrelationships to produce this AI-training tool. In collaboration with IBM, the Dundee team is building Watson's familiarity with webs of human reasoning.
View Full Article - May Require Free Registration | Return to Headlines | Share Facebook  LinkedIn  Twitter 

Are Fitness Trackers Fit for Security?
Technical University of Darmstadt (Germany) (09/09/16)

Researchers at the Technical University of Darmstadt (TU Darmstadt) in Germany conducted a study investigating fraud opportunities within fitness trackers and detected serious security vulnerabilities. Many trackers monitor distances run, measure heart rate and pulse, and check if the user is asleep, but "these data are not only used for the original purpose but are increasingly being used by third parties," says TU Darmstadt professor Ahmad-Reza Sadeghi. He notes data collected by fitness trackers has been used as evidence in court trials, and some health insurance companies recently began offering discounts if insured persons provide personal data from their trackers. Sadeghi says these practices could attract scammers who manipulate the tracked data to fraudulently gain financial benefits or influence a court trial. The researchers investigated the security of fitness trackers by manipulating the data on their way to the cloud server using a man-in-the-middle attack and examined the security of communication protocols used by the fitness trackers. They found although all cloud-based tracking systems use an encrypted protocol to transfer data, they were able to falsify data. The flaws could be corrected with known standard technologies, but Sadeghi says, "the manufacturers have to put some more effort in employing these technologies in their products."

Virginia Tech Initiative With Qualcomm to Expose Students to STEM
Washington Post (09/11/16) Shapiro; T. Rees

Virginia Polytechnic Institute and State University (Virginia Tech) and computer chip manufacturer Qualcomm have teamed up to open a new Thinkabit lab at the university's Northern Virginia campus, aimed at exposing students to science, technology, engineering, and math (STEM) fields. "The important thing for students coming out here is that students learn there is a place for them in the world," says Qualcomm's Susie Armstrong. Qualcomm's first Thinkabit lab opened in San Diego more than a year ago and has served 8,000 students around California across four other locations. The Virginia Tech lab is free for community groups and all school districts in the Washington, D.C., area. Over the course of the six-hour program, students will learn about STEM career opportunities and engage in various activities and projects. Virginia Tech president Timothy Sands had pushed for a program to introduce students as young as middle-schoolers to STEM as part of a broader effort to prepare the university for the evolving needs of the workforce. "The jobs that are going to be growing is the intersection of social science and technology," Sands says. "But the thing about technology is it changes rapidly."
View Full Article - May Require Free Registration | Return to Headlines | Share Facebook  LinkedIn  Twitter 

$4.6 Million Grant to Improve How Automated Cars, Drones Interact With Humans
Berkeley News (09/08/16) Robert Sanders

University of California, Berkeley (UC Berkeley) researchers are using a five-year, $4.6-million grant from the U.S. National Science Foundation to improve interaction between humans and unmanned, autonomous vehicles. The researchers have founded the Verified Human Interfaces, Control, and Learning for Semi-Autonomous Systems (VeHICaL) group to develop and confirm the design of human-machine interfaces that govern cyber-physical systems. VeHICaL leader and UC Berkeley professor Sanjit Seshia says the project seeks to more deeply understand how humans and machines can work in unison to perform safety- and mission-critical tasks in society. "As intelligent cyber-physical systems are deployed in critical sectors such as transportation, aerospace, and healthcare, there is a pressing need to design for their interaction with humans so as to ensure that safety, security, privacy, and performance objectives are met," Seshia notes. The VeHICaL group will tap new interdisciplinary research combining concepts in formal methods, control theory, robotics and perception, cognitive science, machine learning, security and privacy, and human-machine interfaces. The initiative will produce theory and tools for designers of human-cyber-physical systems to develop next-generation verified intelligent systems that collaborate with people to execute complex tasks with demonstrable assurances on safety, privacy, and performance.

CMU Algorithm Detects Online Fraudsters
Carnegie Mellon News (PA) (09/08/16) Byron Spice

Carnegie Mellon University (CMU) researchers say they have developed an algorithm called FRAUDAR that can perceive fraudsters hiding behind a digital veneer of legitimacy. CMU professor Christos Faloutsos says FRAUDAR analyzed Twitter data for 41.7 million users and 1.47 billion followers to detect more than 4,000 accounts not previously tagged as fraudulent, including many that employed known follower-buying services. "We're not identifying anything criminal here, but these sorts of frauds can undermine people's faith in online reviews and behaviors," Faloutsos says. FRAUDAR first locates accounts it can confidently rate as legitimate and then eliminates them until it drills down to a bipartite core. The researchers randomly chose 125 followers and 125 followees from the suspect group, and two control groups of 100 users who had not been picked out by FRAUDAR. They examined each for links associated with malware or scams and for robot-like behavior, and found 57 percent of followers and 40 percent of followees in the suspicious group were labeled as fraudulent, versus 12 percent and 25 percent, respectively, in the control groups. A paper detailing FRAUDAR won the Best Paper Award in August at the ACM Conference on Knowledge Discovery and Data Mining (KDD 2016) in San Francisco.

New Service Improves Cloud Storage Usage on Mobile Devices
Binghamton University (09/07/16)

Researchers at Binghamton University have developed StoArranger, a new service designed to improve the user experience for those saving data from mobile devices to the cloud. StoArranger is designed to intercept, coordinate, and optimize requests made by mobile applications and cloud storage services, according to the researchers. They say the way apps or an iPhone or Android-based device runs does not change because StoArranger works as a "middleware system." StoArranger intercepts cloud storage requests and orders them in the best way to save power, complete tasks as quickly as possible, and minimize the amount of data used to complete the tasks. "We are trying to solve problems without changing operating systems or the existing apps, which makes our solution practical and scalable to existing smartphone users," says Binghamton professor Yifan Zhang. The team plans to develop an app for public use. The researchers presented a paper about their work in August at the ACM SIGOPS Asia-Pacific Workshop on Systems (APSys 2016) in Hong Kong. "The programming committee thought the work presented is a good demonstration of the negative effects of the way that current cloud storage providers chose to deploy their services," Zhang says. "The solution we proposed could be a practical way to solve the problem."

Smartphone Hacks 3D Printer by Measuring 'Leaked' Energy and Acoustic Waves
UB News Center (09/07/16) Cory Nealon

State University of New York at Buffalo (UB) researchers say anyone with a smartphone could potentially steal intellectual property from a business' three-dimensional (3D) printer. Although analysts say 3D printing will become a multibillion-dollar industry, the technology's security vulnerabilities have not been addressed. UB researchers programmed a smartphone's sensors to measure the electromagnetic energy and acoustic waves produced by 3D printers, and used this data to infer the location of the print nozzle as it created the printed object. At 20 centimeters away from the printer, the researchers found the phone was able to gather enough data to enable them to replicate a simple printed object with a 94-percent accuracy rate. "Smartphones are so common that industries may let their guard down, thus creating a situation where intellectual property is ripe for theft," says UB professor Chi Zhou. The researchers say there are several ways to make 3D printing more secure, including restricting access to a machine printing sensitive materials, and increasing the speed of printing. They also propose software-based solutions, such as programming the printer to operate at different speeds, and hardware-based concepts, such as acoustic and electromagnetic shields.

Introducing Diversity in Online Language Analysis
UMass Amherst (09/07/16) Janet Lathrop

Researchers at the University of Massachusetts, Amherst (UMass Amherst) collaborated to broaden the use of natural language processing (NLP) by training computers to identify words, phrases, and language patterns associated with African-American English. They analyzed dialects found on Twitter used by African Americans, and identified these users with U.S. census data and Twitter's geolocation features to correlate African-American communities via a statistical model based on a soft correlation between demographics and language. The researchers validated the model by checking it against insight from earlier linguistics research, and found it can successfully recognize patterns of African-American English. Lisa Green at UMass Amherst's Center for Study of African-American Language says they also uncovered "new phenomena that are not well known in the literature, such as abbreviations and acronyms used on Twitter, particularly those used by African-American speakers." The last stage was to compare the model to existing language classifiers to ascertain how well those NLP tools perform in examining African-American English in user-level and message-level analyses. UMass Amherst professor Brendan O'Connor says popular tools classify African-American English as "not English" at higher rates than expected. He also notes these findings have a direct bearing on "the issue of fairness and equity in artificial intelligence methods."

Social Networks Enable Smart Household Appliances to Make Better Recommendations
University of the Basque Country (Spain) (09/07/16)

A doctoral thesis by David Nunez at Spain's University of the Basque Country details how the relationship between users and their smart domestic appliances can be improved with social networks. He says the thesis falls within the framework of a European initiative partly concentrating on user engagement with smart domestic appliances connected to a smart module. Nunez covers three lines of research in social networks, with the first seeking to anticipate the trust a user will place in another belonging to their social setting on the basis of other contacts' opinions about the target user. Nunez says he developed algebra-based trust prediction tools that are more straightforward than literary ones. He also conducted experiments on recommendation systems, proposing methods based on the Web of Trust of the target user to whom one intends to make a recommendation and on similarities between users and the means of assessment they have. Nunez says the third line of research on maximizing influence led to "a new algorithm that improves the algorithm that exists in the literature in terms of time: the classical Greedy method. Our method has succeeded in getting closer to the optimum like the Greedy one, but does so more rapidly."

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

Correction: In Friday's edition of TechNews, in the article "Montreal Universities Land Historic $213M Investment for Computer and Brain Research," we mistakenly identified researcher Yoshua Bengio as being from McGill University; he actually is associated with University of Montreal. We regret the error.

To submit feedback about ACM TechNews, contact: