#безопасность
-
Ex-NSA chief: water utility controllers don't belong on the internet
After a series of suspected Iranian attacks, the former head of the US National Security Agency has called for industrial water-system controllers to be pulled off the open internet.
-
Zero-Day in Metabase: SQL Injection with CVSS 10.0 Used to Steal Customer Data
A critical vulnerability in the Metabase analytics platform allowed an unauthenticated attacker to gain administrator privileges. Among those affected are laptop maker Framework and the form service Tally.
-
Levi Strauss Says Corporate Data Was Stolen After Attack on Employees
Using social engineering, hackers broke into the work computers of three employees of the jeans maker and stole corporate information. According to the company, customer data was not affected.
-
ICE Is Buying Credit-Card Application Data from Brokers
According to Bruce Schneier, the U.S. immigration service is obtaining, through intermediaries, information that people supplied when applying for credit cards.
-
HTTP Terminator: PortSwigger's AI System Finds Novel HTTP Desync Techniques and a 0-Day in Apache
James Kettle's research system churned through 30,000 attack vectors to surface fresh HTTP desync tricks, while a parallel investigation laid bare a zero-day in Apache Traffic Server.
-
A Single GitHub Issue Reached the CI Secrets of Claude Code and Gemini CLI
Researchers at Novee Security showed at Black Hat USA that an issue filed from an account with no repository privileges can execute code on the CI runners of Anthropic's, Google's and OpenAI's own agents.
-
The Government of Nepal Joins Have I Been Pwned
Nepal's National Cyber Security Centre has become the 47th government agency to receive free access to Troy Hunt's service for monitoring data breaches.
-
ClickFix for macOS: an infostealer quietly skims a slice of your crypto
Researchers at Huntress have uncovered a Go-based piece of malware that steals passwords, Apple Keychain data and hijacks cryptocurrency transactions through fake instructions in Terminal — but drains only a portion of the wallet rather than the whole thing.
-
LightSpy Spyware Operator Undone by a KFC Order in His Real Name
Arctic Wolf researchers tied the Chinese LightSpy spyware to a commercial contractor after one of its operators ordered food from KFC using his real name and office address.
-
Interrupt Injection: A Precisely Timed Interrupt Slips Past Spectre v2 Defenses
MIT CSAIL researchers have shown that an unprivileged program on Linux can wedge itself into the gap between the moment the branch predictor is cleared and the moment the kernel uses it — re-“poisoning” it and bypassing the standard mitigations against Spectre v2.
-
Weak Random Number Generator in CryptoJS Led to $5.7M in Crypto Wallet Thefts
The firm Coinspect has determined that the source of a string of crypto wallet thefts was the CryptoJS.lib.WordArray.random() function, which produced predictable seed phrases. Five applications were affected, and measured losses topped $5.7 million.
-
“Ask AI”: How Website Buttons Quietly Poison Chatbot Memory
A new kind of prompt injection needs neither malicious code nor stolen passwords. It abuses a standard feature of every major AI assistant — prefilled links.
-
Over 4,400 Rockwell controllers exposed online, 22 in the towns where water utilities were attacked
Forescout counted more than four thousand internet-accessible Rockwell Automation PLCs, some of them located in the towns where U.S. water utilities were recently hit. The researchers found no confirmed breaches among them.
-
181,874 meetings wide open: how you could read other people's calls in tl;dv
A security researcher discovered that tl;dv, an AI meeting-recording platform, kept its Firestore database with no isolation between accounts for months. Any free-tier user could enumerate every meeting on the service — and even barge into someone else's live call.
-
CoreBreak: AWS, Google and Vercel Agents Tricked Into Running Tools Without the Model
At Black Hat USA 2026, researchers from Stealth demonstrated a general class of attacks in which forged tool calls reach an AI agent's execution layer without the language model ever taking a turn. AWS, Google and Vercel have shipped fixes.
-
AI agents created fake profiles for a cyberattack in AISI tests
The UK's AI Safety Institute found that Anthropic and OpenAI models resorted to deception on their own during trials, creating fake accounts of real people and attempting to slip malicious code into GitHub.
-
Zbtlink Routers Ship with a Factory Backdoor That Opens Passwordless Root Access
Researchers at VulnCheck have found a built-in backdoor in every firmware image from Chinese manufacturer Zbtlink: it hands an attacker an interactive root shell without any authentication whatsoever.
-
A Meta model accidentally broke into someone else's system during testing — the third such case in a row
Meta has confirmed that its Muse Spark AI model penetrated a third party's system during trials. Anthropic and OpenAI had earlier admitted to similar incidents.
-
Thousands of servers are vulnerable through their motherboard management controllers
At the Black Hat conference, researcher HD Moore showed that the BMC controllers built into server boards remain an “under-monitored” parallel attack surface. More than half of the devices found online contain critical vulnerabilities.
-
Hackers Hid the khunt Toolkit Inside an Oracle Database
Through an SQL injection, attackers planted a post-exploitation kit directly inside an Oracle database, compiling it as a Java object. The technique, according to Huntress, is spotted extremely rarely in real-world attacks.
- More