Hi Cardinal readers, and thanks for scoping the latest edition of Tech Briefs, a weekly batch of items covering the digital and life sciences landscapes. The roundup goes live on Wednesdays.
Got tips and/or questions? Reach out to me via tad@cardinalnews.org.
What’s inside:
Hackers’ work stealing, exploiting images is cheaper, simpler in generative AI era
There is no guarantee that a hacker with store-bought artificial intelligence models and a knowledge of simple commands can’t snag your publicly posted images for a myriad of unauthorized uses, a research team has found.
The team, led by Virginia Tech cybersecurity expert Bimal Viswanath, delivered findings in March during the Institute of Electrical and Electronics Engineers’ Conference on Secure and Trustworthy Machine Learning in Munich, Germany, according to a university news release.
Once images are acquired, hackers can use them for artificial intelligence training, style mimicry and deepfake manipulations, according to the release.
“We stress the urgent need to develop robust defenses and establish that any future protection mechanism must be benchmarked against attacks from off-the-shelf GenAI models,” Viswanath wrote with a Virginia Tech colleague, assistant professor Peng Gao; doctoral students Xavier Pleimling and Sifat Muhammad Abdullah; a University of Texas at San Antonio professor; and two professors from the Indian Institute of Technology, Kharagpur.
According to the paper’s abstract, generative AI model advances have fostered the development of multiple strategies to protect images from unauthorized use. Those strategies employ so-called “protective perturbations” — imperceptible sounds embedded onto images to thwart models from “learning” and misusing them. Previously, attackers required made-to-order methods to neutralize such protections, but now someone can give a text prompt to an “off-the-shelf image-to-image GenAI” model to “denoise” images.
In eight case studies with six diverse protection schemes, the team’s general-purpose attacks breached the noises better than more specialized attacks “while preserving the image’s utility for the adversary,” the documents read. “Our findings reveal a critical and widespread vulnerability in the current landscape of image protection, indicating that many schemes provide a false sense of security.”
Read the entire paper online.
“As GenAI models continue to grow in capability, this threat will only become more severe,” the team wrote. “Therefore, we stress the urgent need for the research community to develop a new generation of robust protection schemes. We posit that resilience against this simple denoising attack must serve as a fundamental benchmark for any future defense mechanism.”
Entrepreneur Resource Center opens in Blacksburg
Virginia Tech Corporate Research Center leaders cut the ribbon Tuesday for a new hub focused on business founders in Blacksburg.
The Entrepreneur Resource Center, ERC for short, was designed to change how founders start, navigate and grow their businesses, VTCRC spokeswoman Danielle Akers said in a recent email.
“At a time when entrepreneurs are often navigating fragmented resources and unclear pathways, the ERC creates something different: a structured, accessible entry point into the innovation ecosystem,” Akers wrote, describing it as one “place where ideas can gain traction, where founders can connect to the right people faster, and where momentum can turn into real companies.”
Programming, mentorship and connections to capital will be available at ERC, which is open to entrepreneurs at every stage, including faculty innovators and Virginia Tech alumni and students, as well as founders from the region, according to its website.
Get more information, including a link to participate, at vtcrc.com/entrepreneur-resource-center.
Second state-funding round set for EV charging assistance
The Virginia Department of Energy has announced a new round of funding for electric vehicle charging infrastructure in the commonwealth’s underserved communities, according to a department news release.
Applications are due June 30 to participate in a program designed to broaden access to EV supply equipment, fill gaps in charging availability and deliver community benefits meant to broaden EV adoption.
Virginia Energy will make up to $510,000 available to applicants that include for-profit businesses and government-industry partnerships to purchase and install EV charging equipment in projects expected to continue through the end of 2027, according to the news release.
This is the second round of Electric Vehicle Charging Assistance Program funding, focusing on historically underserved areas, including financially and environmentally disadvantaged communities, brownfield sites and previously mined areas.
“Expanding EV charging access in underserved communities is essential to ensuring that Virginia’s transportation future is reliable, practical, and accessible to everyone,” said Michael Skiffington, Virginia Energy’s acting director, in the news release. “Through this funding opportunity, Virginia Energy is working to support projects that meet community needs, strengthen local infrastructure, and help more Virginians benefit from the transition to cleaner transportation options.”
According to the release, funded projects would provide data on charging usage, operational performance, user accessibility, and economic and community impacts while complying with federal requirements.
Deadline for applications through the Virginia Energy Grant Portal is due by 11:59 p.m. June 30. Questions submitted through the portal’s help questions tab are due by June 23.
