Welcome to the November 14, 2022, 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).
To view "Headlines At A Glance," hit the link labeled "Click here to view this online" found at the top of the page in the html version.
The online version now has a button at the top labeled "Show Headlines."
|
|
App Identifies Parkinson's, COVID-19 Based on User's Voice
IEEE Spectrum Michelle Hampson November 8, 2022
Machine learning algorithms incorporated into a smartphone application analyze people's voices to identify those in the early stages of Parkinson's disease (PD) or severe COVID-19 lung infections. Researchers at Australia's Royal Melbourne Institute of Technology (RMIT) recorded on an iOS-based smartphone study participants uttering different phonemes that required sounds from the throat, the mouth, and the nose. then created and applied an algorithm to differentiate between persons with PD and healthy volunteers. The algorithm could identify PD sufferers with 100% accuracy. A different algorithm co-developed by RMIT researchers was able to phonemically differentiate patients with COVID-19 lung infections from healthy controls in Indonesia with 94% accuracy.
|
Another Reason to Hate Unwanted Ads
Georgia Tech News Center November 10, 2022
Researchers from the Georgia Institute of Technology (Georgia Tech), University of Illinois Chicago, and New York University found that the process used by third-party advertisers to target online users can be viewed or manipulated using a target's email address. They found that once a user’s email address is uncovered, the information being collected by any third-party advertiser observing a specific user’s targeted ad stream can be tapped into, allowing insight into an individual’s browsing history. Said Paul Pearce of Georgia Tech, “Our work shows the way that information is passed to the ad networks is both insecure and hard to verify. If an attacker knows a victim’s email address, they can lie to the ad network pretending to be a user, leading to very real privacy problems.”
|
Improving Security for Smart Systems
WSU Insider Tina Hilding November 7, 2022
Washington State University (WSU) researchers have developed a statistical analysis technique for complex sensor data that can strengthen decision-making algorithms' resilience and error tolerance. Hackers can cause small perturbations in smart sensors' data that human monitors overlook, leading to prediction and decision-making failures. The WSU researchers enhanced their algorithm with a security layer that can prevent failures by looking for potential disturbances and determining their statistical likelihood. They used the algorithm with health-monitoring wearables to account for actual data disturbances, improving accuracy by 50% compared to standard machine learning algorithms that need clean data. WSU's Jana Doppa called the achievement "an important and novel contribution in the area of security of machine learning systems."
|
Researchers Develop Meta-Reinforcement Learning Algorithm for Traffic Signal Control
Chung-Ang University (South Korea) November 9, 2022
A meta-reinforcement learning (RL) model for traffic signal control developed by researchers at South Korea's Chung-Ang University can adjust its goal based on the traffic environment. The extended deep Q-network-incorporated context-based meta-RL model uses a latent variable that indicates the overall environmental condition to assess traffic flow, then implements traffic signal phases to either maximize throughput or minimize delays. The action is controlled via a "reward," with the reward function set corresponding to a better or worse performance in managing traffic relative to the prior interval. Based on tests using a commercial traffic simulator and real-world tests at 15 intersections in Seoul, the researchers found the model could switch control tasks without explicit traffic information and differentiate between rewards in accordance with traffic saturation levels.
|
VR Remote Collaboration System Lets Users Share Experience on the Move Without Causing VR Sickness
News-Medical.Net November 6, 2022
A virtual reality (VR) remote collaboration system developed by researchers at Japan's Tokyo Metropolitan University aims to reduce VR sickness in immersive, three-dimensional (3D) environments. The system allows users on Segways equipped with 3D cameras and accelerometers to share their observations and the feeling of movement in real time with remote users in modified wheelchairs wearing VR headsets. Tests found that the addition of movement resulted in a 54% reduction in VR sickness, which occurs when users see motion in VR headsets without actually moving. The researchers found the optimal setting involved sending about 60% of the acceleration suggested by visual cues to the modified wheelchair's wheels.
|
En Route to Human-Environment Interaction Technology with Soft Microfingers
Ritsumeikan University (Japan) November 8, 2022
Researchers at Japan's Ritsumeikan University (Rits) have developed a soft robotic microfinger that enables direct human-insect interaction. Rits' Satoshi Konishi said the tactile microfingers are formed from liquid-metal flexible strain sensors manipulated by soft pneumatic balloon actuators. Users control the microfingers with a robotic glove. The researchers used the technology to test the reaction force of pill bugs, which was measured from the insect's legs at roughly 10 millinewtons. Said Konishi, "We anticipate that our results will lead to further technological development for microfinger-insect interactions, leading to human-environment interactions at much smaller scales."
|
Ransomware Gangs Shift Tactics, Making Crimes Harder to Track
Bloomberg Jack Gillum November 11, 2022
Research by Recorded Future Inc.'s Allan Liska found that more ransomware gangs are using their own or stolen computer code in an effort to make it harder to monitor their activity. Liska said, "In the last year, ransomware has become a race to bottom among ransomware groups," with gangs "stealing from each other, lying even more than usual to victims, and creating havoc among investigators and law enforcement.” This comes amid an increase in the number of smaller hacking groups, which Liska said may be concerned about being targeted as part of a larger group. Recently, hackers tied to the Netwalker and REvil extortion groups pleaded guilty, and a dual Russian and Canadian national was charged on allegations of working with the LockBit ransomware gang.
*May Require Paid Registration
|
Go-Playing Trick Defeats World-Class Go AI—but Loses to Human Amateurs
Ars Technica Benj Edwards November 7, 2022
University of California, Berkeley (UC Berkeley) researchers used a neural network and the Monte-Carlo Tree Search method to identify moves in the ancient board game Go that can defeat open source artificial intelligence KataGo by exploiting its blind spots. The weaker adversarial Go-playing program can trick KataGo into losing by playing unexpected moves outside KataGo's training set. The adversarial policy first claims a small corner of the board, allowing KataGo to claim the remainder of the board. UC Berkeley's Adam Gleave said, "This tricks KataGo into thinking it's already won, since its territory (bottom-left) is much larger than the adversary's. But the bottom-left territory doesn't actually contribute to its score (only the white stones it has played) because of the presence of black stones there, meaning it's not fully secured." Gleave said that while the result is entertaining in Go, “Similar failures in safety-critical systems could be dangerous."
|
Contactless Screening Tool Could Revolutionize Chronic Wound Treatment
RMIT University (Australia) November 10, 2022
Researchers at Australia's Royal Melbourne Institute of Technology (RMIT) and Bolton Clarke Research Institute have demonstrated that a thermal-imaging tool can identify chronic leg wounds during an initial home assessment. This builds on previous research by the team in which chronic leg wounds were identified by the second week after a baseline assessment. Powered by artificial intelligence, the tool uses thermal images captured during a first assessment to predict how these wounds will heal. By assessing spatial heat distribution in a wound, the tool was 78% accurate in determining whether venous leg ulcers in 56 patients would heal in 12 weeks without specialized treatment. When using the new tool, said RMIT's Dinesh Kumar, "Specialized treatment for slow-healing leg ulcers can begin up to four weeks earlier than the current gold standard."
|
U.K. Will Use GPS Fingerprint Scanner to Track People Facing Deportation
New Scientist Jason Arunn Murugesu November 11, 2022
The U.K.'s Home Office says it will soon require people facing deportation to carry a global positioning system (GPS)-enabled fingerprint scanner, so the Home Office can determine their location and identify at all times. Adult foreign nationals subject to deportation orders will be required to scan their fingers when prompted on the key fob-like devices from Buddi to verify their identity and proximity to the device. Lucie Audibert at Privacy International says the scanners, which will launch this fall, will be just as intrusive as ankle tags. "It may also feed into the normalization of GPS tracking as it becomes physically and morally more tolerable and acceptable to wear this new device than an ankle tag that is loaded with stigma," Audibert warned.
|
AI Deciphers Detector 'Clouds' to Accelerate Materials Research
SLAC National Accelerator Laboratory Chris Patrick November 7, 2022
A team of scientists led by Joshua Turner at the U.S. Department of Energy's Stanford Linear Accelerator Center National Accelerator Laboratory has developed an algorithm that extracts data to accelerate materials research with X-ray pulse pairs. The machine learning technique uses raw detector imagery of scattered photons to extract fluctuation information 10 times faster on its own and 100 times faster in conjunction with upgraded hardware, enabling near-real-time data analysis. This resolves charge clouds formed when speckle patterns of scattered photons merge. The algorithm learned how the charge clouds merge, and was able to unsnarl how many photons struck the detector per blob and per pulse pair.
|
Blockchain Game Could Help Create Metaverse No One Owns
MIT Technology Review Mike Orcutt November 10, 2022
The sci-fi-themed online game Dark Forest operates on a blockchain, meaning no one can manipulate its outcome. Dark Forest was conceived by pseudonymous programmer "Gubsheep," who characterizes it as a "massively multiplayer strategy game that takes place in an infinite, procedurally generated universe." The game uses cryptographic zero-knowledge proofs to hide opposing players from each other as they engage in empire-building. New players are confronted with a mostly hidden universe that becomes visible through exploration; when players move, they send a validating proof to the blockchain without exposing their coordinates. Some players envision Dark Forest as the first step toward metaverses driven by decentralized networks rather than company servers.
|
|