UAT-11985 targets Taiwan research affiliates with fake event invites that lead to AitM pages designed to capture Google passwords and MFA codes.
Get the latest international news and world events from around the world.
Hackers exploit Collabora RCE flaw to backdoor Nextcloud servers
Threat actors are exploiting a critical command execution flaw, now tracked as CVE-2026–110447, in the Collabora Online — Built-in CODE Server app (richdocumentscode) for Nextcloud.
Multiple administrator reports indicate that attacks began in mid-September, roughly a month before the formal security bulletin, the GitHub advisory, and CVE publication.
CVE-2026–110447 impacts richdocumentscode versions before 26.4.303 and resides in the Nextcloud app’s mechanism for starting the bundled Collabora server.
Targeted ads installed Candiru spyware in zero-click attacks
Candiru spyware was delivered to devices at Al Jazeera Media Network via targeted online advertisements in a zero-click attack that required only rendering the malicious ads on the victim’s device.
Researchers at Citizen Labs investigating the attacks referred to this technique as AdFection, noting that their report is the first public technical confirmation of mercenary spyware delivered through targeted ads.
A similar method was documented in late 2025 by Amnesty International, involving ads that infected targets with Predator spyware without them clicking anything.
NASA to Help Unlock Mysteries of Mars’ Moons in Upcoming Mission
NASA is playing an important role in JAXA’s (Japan Aerospace Exploration Agency) MMX (Martian Moons eXploration) mission, to advance our understanding of Mars’ moons and build on the agency’s long history of exploring the Red Planet.
The mission aims to characterize and determine the origin of Mars’ moons Phobos and Deimos and deliver a sample of Phobos to Earth. The MMX spacecraft is targeting launch from JAXA’s Tanegashima Space Center in Japan at 3:41 p.m. EDT on Monday, Oct. 19 (4:41 a.m. JST on Tuesday, Oct. 20).
Anthropic is “destructively scanning” millions of books. Should you care?
An informative data-driven perspective which explores why the alarming-sounding headlines about Anthropic’s destructive book scanning efforts may be misleading (and even manipulative). To be clear, I’m not saying that all critiques of tech companies are invalid. It’s just that the evidence for the validity of this specific moral panic is thinner than it might appear.
I’m a proponent of looking deeply into the underlying context and data of complex issues like this one. Many news outlets do report in a negativity-biased fashion since they can stir up outrage and get more attention. So, I feel it is important to be media literate and pay attention to the nuances of issues.
I should also note that this is just a Substack post and thus does not constitute a verified primary source like a scientific paper. From what I can tell, it makes legitimate arguments and presents legitimate data, but this is not a 100% certain conclusion.
On the curious debate over AI companies buying, digitizing, and discarding “rare” books.
A protein that fuels cancer could also be the key to fighting it
The findings, published in Nature Immunology, could be especially important for colorectal cancer, which has become the leading cause of cancer-related death among people under 50. Many patients with difficult-to-treat forms of the disease have limited options, particularly when chemotherapy and newer immunotherapy treatments fail. The research suggests that a protein normally associated with cancer growth could instead become a powerful cancer-fighting “force” when activated in the right cells.
Turning a Cancer-Promoting Protein Into an Immune System Weapon
Scientists have long understood that a patient’s physical fitness can influence their ability to withstand illness. But CSHL Assistant Professor Semir Beyaz is exploring a different kind of fitness: the metabolic health and performance of individual cells.
Application of Artificial Intelligence in the Prediction and Management of Stroke Rehabilitation
Stroke prediction research includes acute diagnosis, acute prognostication, rehabilitation-outcome forecasting, and long-term risk management. Each addresses a different clinical question and requires different supporting evidence. Even technically sophisticated models may be difficult to use across heterogeneous settings because performance in a development dataset may not persist in other populations or fit routine workflows.
Acute imaging and early prognostic models provide baseline information that may inform subsequent rehabilitation planning. Neuroimaging can characterize lesion location, volume, and tissue viability; early clinical severity scores capture initial impairment; and acute-treatment data document reperfusion status and complication risk. These variables may be incorporated as covariates in rehabilitation prediction models, while timely acute diagnosis and treatment may influence the functional baseline from which rehabilitation begins.
Measures of acute diagnostic accuracy or early prognostic discrimination answer an early-care question; they do not show that a rehabilitation intervention is effective or that algorithm-guided adjustment improves long-term recovery. Rehabilitation claims therefore require models developed and validated for outcomes such as motor function, ADLs, participation, or quality of life. Whenever possible, studies should also test whether using the model changes clinical decisions or patient outcomes compared with standard care.
Cancer Predisposition Syndromes in Children: Who, When, and How to Screen?
Anupindi SA, Bedoya MA, Lindell RB, et al. Diagnostic Performance of Whole-Body MRI as a Tool for Cancer Screening in Children with Genetic Cancer-Predisposing Conditions. AJR Am J Roentgenol 2015;205:400–408.
Identifying Genomic Biomarkers of Rare Disease Subtypes Using Minority-Weighted Feature Selection
Introduces a novel minority-aware and robust feature selection framework, GenMWS, for high-dimensional class-imbalanced gene expression data.
Constructs class-specific core intervals and weighted own-class/other-class membership scores to capture gene-wise class structure.
Develops a weighted nonoverlap mask and class-specific support summaries to emphasize minority-class informative patterns while controlling majority-supported structure.