Sign in Subscribe

InfoSec News 24FEB2026

General

  • More from Brian Krebs - Starkiller Phishing as a Service. They're using in-URL authentication to trick users into thinking it's the legitimate site (using the legitimate site's domain as the username - hxxp://username:password@site/). It's also proxying the real site, via a headless Chrome browser, so that the site content looks and acts correctly.
    The only real solution - phishing-resistant MFA, such as Passkeys.
Most phishing websites are little more than static copies of login pages for popular online destinations, and they are often quickly taken down by anti-abuse activists and security firms. But a stealthy new phishing-as-a-service offering lets customers sidestep both of these pitfalls: It uses cleverly disguised links to load the target brand’s real website, and then acts as a relay between the victim and the legitimate site — forwarding the victim’s username, password and multi-factor authentication (MFA) code to the legitimate site and returning its responses.
An attacker enters a brand’s real URL, and the platform spins up a Docker container running a headless Chrome instance that loads the real login page.
The container then acts as a man-in-the-middle reverse proxy, forwarding the end user’s inputs to the legitimate site and returning the site's responses. Every keystroke, form submission, and session token passes through attacker-controlled infrastructure and is logged along the way.
When using classic Outlook, you may find that the mouse pointer or mouse cursor disappears as you move the pointer over the Outlook interface. Although the mouse pointer is not there, the email in the message list will change color as you hover over it. This issue has also been reported with OneNote and other Microsoft 365 apps to a lesser degree.
...
To work around the issue, try the following:
...
3. Restart your computer.

Getting Techy

  • Something a little different - Trail of Bits applying their TRAIL (Threat and Risk Analysis Informed Lifecycle) threat-modelling approach to Perplexity's Comet (Agentic AI) browser.
    Interesting aside, when performing prompt-injection attacks:
The misspellings (“browisng,” “succeeidng,” “existnece”) were accidental typos in our initial proof of concept. When we corrected them, the agent correctly identified the warning as fraudulent and did not act on it. Surprisingly, the typos are necessary for the exploit to function.

Geo-Politics

  • Politico conducted a poll across Canada, France, Germany, UK and US, covering cyber-attacks, and government responses to the attacks.
NATO countries’ restrained response to hybrid attacks is at odds with public opinion, new polling shows: Broad swaths of the public in key allied countries say actions such as cyberattacks on hospitals should be considered acts of war.
...
State-backed hackers — often linked to Russia — have increasingly targeted critical sectors in recent years. But NATO allies are struggling to respond effectively.
...
Security services in Europe have also more firmly called out the Kremlin for orchestrating digital attacks in the West, most recently targeting Poland’s energy infrastructure. But views on Russia as a global threat vary greatly between Europe and North America. A majority of respondents in Germany, France and the U.K. said Russia represents the biggest threat to peace, while fewer in the U.S. (39 percent) and Canada (29 percent) agreed.
...
At least a third of respondents in each country agreed that cybersecurity and defense against cyber attacks should be among their countries’ highest priorities for defense spending.
“Just being resilient alone, you can’t absorb all threats,” Dag Baehr, Vice President of Germany’s federal intelligence service (BND), said at the Munich Cyber Security Conference last week. “You need to be active in defending.”
Recent ransomware attacks targeting Romania’s critical infrastructure were likely part of a broader Russian hybrid operation aimed at undermining the country’s stability, Romania’s top cybersecurity official said.
..
“These attacks are systematic, well-prepared and often coincide with political decisions or social developments in Romania, particularly those linked to support for Ukraine,” Dan Cimpean, head of Romania’s National Cybersecurity Directorate, said on the sidelines of the Kyiv International Cyber Resilience Forum last week.
We’re still eight months out from Election Day, but the cryptocurrency industry has already dumped at least $288 million into the 2026 midterms — more than double what they spent in the entire 2024 cycle, when their $130 million in spending was itself a historic sum that reshaped Congress. That earlier investment bought them Trump, a deregulatory Congress, and the systematic dismantling of financial oversight that they'd spent years demanding. With Republican anxiety mounting as midterms approach, and the crypto industry’s wishlist only partially fulfilled, it seems they’re preparing to spend whatever it takes to finish the job.
...
To put this in perspective: the industry’s flagship super PAC, Fairshake, is now the #5 most-funded PAC in the country. It trails only Trump’s inaugural committee, MAGA Inc, and the RNC and DNC.

Privacy

  • Samsung's (pre-installed) weather app is leaking a unique fingerprint, allowing tracking of users across networks and VPNs.
Samsung devices ship with a weather application that issues periodic HTTP requests to The Weather Company's API (api.weather.com) at fixed intervals. Each request includes a placeid parameter - a 64-character hexadecimal string, consistent with a SHA-256 digest, that maps to a saved location in the user's weather configuration.
The combination of placeid values across a user's saved locations creates a fingerprint that is effectively unique per device, persists across IP address changes, and is trivially observable by the API provider.
...
Static fingerprints. Multiple users maintained an identical placeid set across all observed days.
...
The fingerprint is bound to the device's saved location configuration, not the network session. This means it survives VPN usage, WiFi-to-cellular handoffs, carrier IP reassignment, and network roaming.

AI

  • Anthropic doesn't want its models used for "mass domestic surveillance and fully autonomous weapons", and this may impact its use inside the US military.
The Pentagon is considering severing or scaling back its relationship with Anthropic after months of tense negotiations over how the military can use its Claude model...
The dispute intensified following a U.S. raid targeting former Venezuelan President Nicolás Maduro, in which Claude was used via Palantir Technologies’ AI platform, as first reported by the Wall Street Journal. After the operation, an Anthropic employee contacted a counterpart at Palantir, prompting concern inside the Defense Department that the company might object to certain military uses.
At issue is the Pentagon’s demand that AI labs permit use of their models for “all lawful purposes,” including weapons development and intelligence operations. Anthropic has not agreed, seeking carve-outs around mass domestic surveillance and fully autonomous weapons. Other labs, including OpenAI, Google, and xAI, are reported to have lifted ordinary guardrails for Pentagon work.
The Pentagon’s contract with Anthropic, valued at up to $200 million, is now under review.
We have identified industrial-scale campaigns by three AI laboratories—DeepSeek, Moonshot, and MiniMax—to illicitly extract Claude’s capabilities to improve their own models. These labs generated over 16 million exchanges with Claude through approximately 24,000 fraudulent accounts, in violation of our terms of service and regional access restrictions.
These labs used a technique called “distillation,” which involves training a less capable model on the outputs of a stronger one. Distillation is a widely used and legitimate training method. For example, frontier AI labs routinely distill their own models to create smaller, cheaper versions for their customers. But distillation can also be used for illicit purposes: competitors can use it to acquire powerful capabilities from other labs in a fraction of the time, and at a fraction of the cost, that it would take to develop them independently.
...
Without visibility into these attacks, the apparently rapid advancements made by these labs are incorrectly taken as evidence that export controls are ineffective and able to be circumvented by innovation. In reality, these advancements depend in significant part on capabilities extracted from American models, and executing this extraction at scale requires access to advanced chips. Distillation attacks therefore reinforce the rationale for export controls: restricted chip access limits both direct model training and the scale of illicit distillation.
While AI can bring meaningful benefits for individuals and society, recent developments - particularly AI image and video generation integrated into widely accessible social media platforms - have enabled the creation of non-consensual intimate imagery, defamatory depictions, and other harmful content featuring real individuals. We are especially concerned about potential harms to children and other vulnerable groups, such as cyber-bullying and/or exploitation.
...
The co-signatories remind all organisations developing and using AI content generation systems that such systems must be developed and used in accordance with applicable legal frameworks, including data protection and privacy rules.
We also highlight that the creation of non-consensual intimate imagery can constitute a criminal offence in many jurisdictions.
Whilst specific legal requirements vary by jurisdiction, fundamental principles should guide all organisations developing and using AI content generation systems

Subscribe to Deuxieme RE Banque News

Sign up now to get access to the library of members-only issues.
Jamie Larson
Subscribe