PRESS-NEWS.org - Press Release Distribution
PRESS RELEASES DISTRIBUTION

Brain inspires more robust AI

A new technique to protect sensitive AI-based applications from attackers

Brain inspires more robust AI
2023-09-16
(Press-News.org)

Most artificially intelligent systems are based on neural networks, algorithms inspired by biological neurons found in the brain. These networks can consist of multiple layers, with inputs coming in one side and outputs going out of the other. The outputs can be used to make automatic decisions, for example, in driverless cars. Attacks to mislead a neural network can involve exploiting vulnerabilities in the input layers, but typically only the initial input layer is considered when engineering a defense. For the first time, researchers augmented a neural network’s inner layers with a process involving random noise to improve its resilience.

Artificial intelligence (AI) has become a relatively common thing; chances are you have a smartphone with an AI assistant or you use a search engine powered by AI. While it’s a broad term that can include many different ways to essentially process information and sometimes make decisions, AI systems are often built using artificial neural networks (ANN) analogous to those of the brain. And like the brain, ANNs can sometimes get confused, either by accident or by the deliberate actions of a third party. Think of something like an optical illusion — it might make you feel like you are looking at one thing when you are really looking at another.

The difference between things that confuse an ANN and things that might confuse us, however, is that some visual input could appear perfectly normal, or at least might be understandable to us, but may nevertheless be interpreted as something completely different by an ANN.
A trivial example might be an image-classifying system mistaking a cat for a dog, but a more serious example could be a driverless car mistaking a stop signal for a right-of-way sign. And it’s not just the already controversial example of driverless cars; there are medical diagnostic systems, and many other sensitive applications that take inputs and inform, or even make, decisions that can affect people.

As inputs aren’t necessarily visual, it’s not always easy to analyze why a system might have made a mistake at a glance. Attackers trying to disrupt a system based on ANNs can take advantage of this, subtly altering an anticipated input pattern so that it will be misinterpreted, and the system will behave wrongly, perhaps even problematically. There are some defense techniques for attacks like these, but they have limitations. Recent graduate Jumpei Ukita and Professor Kenichi Ohki from the Department of Physiology at the University of Tokyo Graduate School of Medicine devised and tested a new way to improve ANN defense.

“Neural networks typically comprise layers of virtual neurons. The first layers will often be responsible for analyzing inputs by identifying the elements that correspond to a certain input,” said Ohki. “An attacker might supply an image with artifacts that trick the network into misclassifying it. A typical defense for such an attack might be to deliberately introduce some noise into this first layer. This sounds counterintuitive that it might help, but by doing so, it allows for greater adaptations to a visual scene or other set of inputs. However, this method is not always so effective and we thought we could improve the matter by looking beyond the input layer to further inside the network.”

Ukita and Ohki aren’t just computer scientists. They have also studied the human brain, and this inspired them to use a phenomenon they knew about there in an ANN. This was to add noise not only to the input layer, but to deeper layers as well. This is typically avoided as it’s feared that it will impact the effectiveness of the network under normal conditions. But the duo found this not to be the case, and instead the noise promoted greater adaptability in their test ANN, which reduced its susceptibility to simulated adversarial attacks.

“Our first step was to devise a hypothetical method of attack that strikes deeper than the input layer. Such an attack would need to withstand the resilience of a network with a standard noise defense on its input layer. We call these feature-space adversarial examples,” said Ukita. “These attacks work by supplying an input intentionally far from, rather than near to, the input that an ANN can correctly classify. But the trick is to present subtly misleading artifacts to the deeper layers instead. Once we demonstrated the danger from such an attack, we injected random noise into the deeper hidden layers of the network to boost their adaptability and therefore defensive capability. We are happy to report it works.”

While the new idea does prove robust, the team wishes to develop it further to make it even more effective against anticipated attacks, as well as other kinds of attacks they have not yet tested it against. At present, the defense only works on this specific kind of attack.

“Future attackers might try to consider attacks that can escape the feature-space noise we considered in this research,” said Ukita. “Indeed, attack and defense are two sides of the same coin; it’s an arms race that neither side will back down from, so we need to continually iterate, improve and innovate new ideas in order to protect the systems we use every day.”

###

Journal article: Jumpei Ukita and Kenichi Ohki. “Adversarial attacks and defenses using feature-space stochasticity”, Neural Networks, DOI: 10.1016/j.neunet.2023.08.022

Funding:
This work was supported by Brain Mapping by Integrated Neurotechnologies for Disease Studies (Brain/MINDS) from Japan Agency for Medical Research and Development (AMED) (14533320, JP16dm0207034, JP20dm0207048 to K.O.); CREST-JST (JPMJCR22P1 to K.O.); Institute for AI and Beyond (to K.O.); JSPS KAKENHI (25221001, 19H05642, 20H05917, to K.O.); Takeda Science Foundation (to J.U.); and Masayoshi Son Foundation (to J.U.).

Departmental links:
Ohki Lab - https://physiol1.m.u-tokyo.ac.jp/ern24596/en/
Graduate School of Medicine - https://www.m.u-tokyo.ac.jp/english/
International Research Center for Neurointelligence – https://ircn.jp/en/
Institute for AI and Beyond - https://beyondai.jp/?lang=en

Research contact:
Professor Kenichi Ohki
Department of Physiology, Graduate School of Medicine,
The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan
kohki@m.u-tokyo.ac.jp

Press contact:
Mr. Rohan Mehra
Public Relations Group, The University of Tokyo,
7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan
press-releases.adm@gs.mail.u-tokyo.ac.jp

About the University of Tokyo:
The University of Tokyo is Japan's leading university and one of the world's top research universities. The vast research output of some 6,000 researchers is published in the world's top journals across the arts and sciences. Our vibrant student body of around 15,000 undergraduate and 15,000 graduate students includes over 4,000 international students. Find out more at www.u-tokyo.ac.jp/en/ or follow us on Twitter at @UTokyo_News_en.

END


[Attachments] See images for this press release:
Brain inspires more robust AI Brain inspires more robust AI 2

ELSE PRESS RELEASES FROM THIS DATE:

Syphilis transmission networks and antimicrobial resistance in England uncovered using genomics

2023-09-16
Scientists have used genomics to reveal distinct sexual networks for syphilis transmission, defined geographically or by sexual preference, among a background of wider circulation in England. They also show a presence of drug resistance in the majority of cases. By grouping closely related strains of the bacterium that causes syphilis – Treponema pallidum –, researchers demonstrate how a large number of cases are linked together. Researchers from the Wellcome Sanger Institute and their collaborators at the UK Health Security Agency (UKHSA)* sequenced ...

MSU, FRIB developing artificial intelligence tools to enhance discovery, technology and training

2023-09-15
Images Highlights: The U.S. Department of Energy Office of Science, or DOE-SC, is investing in machine learning, a type of artificial intelligence, to accelerate the speed of research and development in nuclear science. Michigan State University researchers at the Facility for Rare Isotope Beams, or FRIB, are leading five of these new grant projects.   These projects aim to enhance the breadth of FRIB’s activities, covering nuclear physics experiments and theory, as well as particle accelerator operations.   FRIB is a DOE-SC user facility, meaning that these advances will serve ...

University of Kentucky Markey Cancer Center attains NCI's highest status as a Comprehensive Cancer Center

University of Kentucky Markey Cancer Center attains NCIs highest status as a Comprehensive Cancer Center
2023-09-15
The University of Kentucky Markey Cancer Center announced today that it has earned a National Cancer Institute “Comprehensive” Cancer Center designation, the highest level of recognition awarded by the NCI. Markey is the first and only center in Kentucky to achieve this designation, and the next-closest Comprehensive Cancer Center is nearly 200 miles from Lexington. There are currently only 72 NCI-Designated Cancer Centers in the country, and 56 of those are Comprehensive Cancer Centers.  As the federal government’s principal ...

Rating platforms drive sales at tourist-area NYC eateries

2023-09-15
Ratings on platforms such as Yelp and TripAdvisor can greatly impact high-priced New York City restaurants that service tourists, but have less of an effect on restaurants frequented by “locals” outside of tourist areas, according to new Cornell research. “In neighborhoods frequented by ‘locals,’ the advent and expansion of internet-based ratings platforms did not result in greater disparities in restaurant sales despite how ubiquitous they are and how frequently we anecdotally use them,” said Jason Greenberg, associate ...

Making AI smarter with an artificial, multisensory integrated neuron

Making AI smarter with an artificial, multisensory integrated neuron
2023-09-15
The feel of a cat’s fur can reveal some information, but seeing the feline provides critical details: is it a housecat or a lion? While the sound of fire crackling may be ambiguous, its scent confirms the burning wood. Our senses synergize to give a comprehensive understanding, particularly when individual signals are subtle. The collective sum of biological inputs can be greater than their individual contributions. Robots tend to follow more straightforward addition, but Penn State researchers have now harnessed the biological concept for application in artificial intelligence (AI) to develop ...

Scientists take next big step in understanding genetics of schizophrenia

2023-09-15
Genetically speaking, we are individuals different from each other because of slight variations in our DNA sequences – so-called genetic variants – some of which have dramatic effects we can see and comprehend, from the color of our eyes to our risk for developing schizophrenia – a debilitating psychiatric condition affecting many millions worldwide. For several years, scientists have studied the entire genomes of thousands of people – called genome-wide association studies, or GWAS – to find approximately 5,000 genetic variants associated with schizophrenia. Now, ...

RIT collaboration with global team confirms, disproves distant galaxies

2023-09-15
Rochester Institute of Technology scientists have once again used data from the James Webb Space Telescope (JWST) as part of the Cosmic Evolution Early Release Science (CEERS) Survey to change the way we think about the universe and its distant galaxies. Jeyhan Kartaltepe, associate professor in the School of Physics and Astronomy, and Rebecca Larson, postdoctoral research associate, co-authored a paper, “Confirmation and refutation of very luminous galaxies in the early Universe,”  published in Nature confirming ...

In major breakthrough, researchers close in on preeclampsia cure

2023-09-15
Researchers from Western and Brown University have made groundbreaking progress towards identifying the root cause and potential therapy for preeclampsia. The pregnancy complication affects up to eight per cent of pregnancies globally and is the leading cause of maternal and fetal mortality due to premature delivery, complications with the placenta and lack of oxygen. The research, led by Drs. Kun Ping Lu and Xiao Zhen Zhou at Western, and Drs. Surendra Sharma and Sukanta Jash at Brown, has identified ...

Facebook's design makes it unable to control misinformation

2023-09-15
WASHINGTON (September 15, 2023) – As misinformation flourished online during the Covid-19 pandemic, a number of platforms announced policies and practices aimed at combating the spread of misinformation. Did those efforts work? New research published today in Science Advances suggests that the Covid-19 vaccine misinformation policies of Facebook, the world’s largest social media platform, were not effective in combating misinformation. The study, led by researchers at the George Washington University, found that Facebook’s efforts were undermined by the core design features ...

Study shows replanting logged forests with diverse mixtures of seedlings accelerates restoration

2023-09-15
Twenty-year experiment finds that active replanting beats natural recovery for restoring logged tropical forests. The higher the diversity of replanted tree species, the more quickly canopy area and biomass recovered. Results emphasize the importance of preserving biodiversity in pristine forests and restoring it in recovering logged forest. Satellite observations of one of the world’s biggest ecological experiments on the island of Borneo have revealed that replanting logged forests with diverse mixtures of seedlings can significantly accelerate their recovery. The results have been published today in the journal Science Advances. The ...

LAST 30 PRESS RELEASES:

12.5, the 1st Impact Factor of COMMTR released!

Circadian clock impact on cluster headaches funded by $2.4M NIH grant for UTHealth Houston research

Study identifies first drug therapy for sleep apnea

How old is your bone marrow?

Boosting biodiversity without hurting local economies

ChatGPT is biased against resumes with credentials that imply a disability — but it can improve

Simple test for flu could improve diagnosis and surveillance

UT Health San Antonio researcher awarded five-year, $2.53 million NIH grant to study alcohol-assisted liver disease

Giving pre-med students hands-on clinical training

CAMH research suggests potential targets for prevention and early identification of psychotic disorders

Mapping the heart to prevent damage caused by a heart attack

Study challenges popular idea that Easter islanders committed ‘ecocide’

Chilling discovery: Study reveals evolution of human cold and menthol sensing protein, offering hope for future non-addictive pain therapies.

Elena Beccalli, new rector of Università Cattolica del Sacro Cuore, takes office on 1st July

Pacific Northwest Research Institute uncovers hidden DNA mechanisms of rare genetic diseases

Empowering older adults: Wearable tech made easier with personalized support

Pennington Biomedical researchers partner on award-winning Long Covid study

Cooling ‘blood oranges’ could make them even healthier – a bonus for consumers

Body image and overall health found important to the sexual health of older gay men, according to new studies

Lab-grown muscles reveal mysteries of rare muscle diseases

Primary hepatic angiosarcoma: Treatment options for a rare tumor

Research finds causal evidence tying cerebral small-vessel disease to Alzheimer’s, dementia

Navigating the Pyrocene: Recent Cell Press papers on managing fire risk

Restoring the Great Salt Lake would have environmental justice as well as ecological benefits

Cannabis, tobacco use, and COVID-19 outcomes

A 5:2 intermittent fasting meal replacement diet and glycemic control for adults with diabetes

Scientists document self-propelling oxygen decline in the oceans

Activating molecular target reverses multiple hallmarks of aging

Cannabis use tied to increased risk of severe COVID-19

How to make ageing a ‘fairer game’ for all wormkind

[Press-News.org] Brain inspires more robust AI
A new technique to protect sensitive AI-based applications from attackers