TLDR InfoSec

TLDR INFORMATION SECURITY 2025-01-28

6 sections17 entrieslinks blockcomplete parse

Parse notes

  • Subscription/account footer truncated before parsefooter_truncated

01GENERAL

Sponsored

PROTECTED ATTACK SURFACE

The web browser is the most widely used enterprise application, but one of the least protected ones. The rise of Highly Evasive Adaptive Threats (HEAT), combined with AI usage, puts browsers in danger of being breached at any moment.

menlosecurity.comSponsor

02ATTACKS & VULNERABILITIES

MILLIONS OF SUBARUS COULD BE REMOTELY UNLOCKED, TRACKED DUE TO SECURITY FLAWS

Security researchers have discovered vulnerabilities in Subaru's Starlink system that allow tracking of vehicles' year-long location history and control of car features. Though now patched, the incident highlighted broader privacy concerns about automotive data collection and employee access to customer information.

arstechnica.com6 min read

BANK OF AMERICA'S SENSITIVE CUSTOMER DATA COMPROMISED IN THIRD-PARTY HACK

Bank of America confirmed that a breach of a third-party software provider led to the exposure of at least 414 customers' personal data. The compromised data includes names, SSNs, addresses, phone numbers, passport numbers, and loan numbers. Bank of America has stated that its systems are not impacted and that it has found no evidence of breach.

benzinga.com1 min read

HUNDREDS OF FAKE REDDIT SITES PUSH LUMMA STEALER MALWARE

Researchers are warning of threat campaigns that are utilizing fake Reddit pages to social engineer users to download files via a fake WeTransfer site. The campaign involves fake Reddit threads where a user asks for help downloading a tool, another user provides a fake WeTransfer link, and then the first user thanks them. The link redirects to a fake WeTransfer page which redirects the user to a page that downloads the Lumma Infostealer malware.

bleepingcomputer.com2 min read

03STRATEGIES & TACTICS

CLONE2LEAK: YOUR GIT CREDENTIALS BELONG TO US

A security engineer investigating GitHub Desktop discovered a vulnerability in Git's Credential Protocol that could leak user credentials. Further research revealed similar bugs across multiple Git-related projects, stemming from improper message handling in credential helper implementations.

flatt.tech11 min read

GETTING STARTED WITH AWS SECURITY: ATTACKING AND AUDITING

A blog post with resources and a lab environment for beginners to get started with AWS security. The post begins with an overview and demonstration of 4 common attacks against EC2 and S3. The post then provides an overview of CLI and GUI based auditing.

cc-connected.com5 min read

04LAUNCHES & TOOLS

AXOFLOW (PRODUCT LAUNCH)

Axoflow automatically discovers security data sources (such as syslog, OpenTelemetry, Windows, and more) and then classifies, normalizes, and enriches the data, reduces data volume by more than 50%, and eliminates the need for manual data wrangling.

axoflow.com

RANDOMPASS

A simple static password generator. Generated passwords are 20 characters, lower and uppercase, and contain special characters by default.

github.comGitHub repository

BAITROUTE

Baitroute is a web honeypot library that creates fake vulnerable endpoints to detect and deceive attackers. It offers easy integration with popular web frameworks and minimal performance impact. By setting up rules and alert handlers, you can monitor and respond to attacker activity on the decoy endpoints.

github.comGitHub repository

05MISCELLANEOUS

BRITISH MUSEUM SAYS EX-CONTRACTOR 'SHUT DOWN' IT SYSTEMS, WREAKED HAVOC

A fired IT contractor broke into the British Museum on January 23 and sabotaged several IT systems, forcing the closure of temporary exhibitions through the weekend. The individual was arrested at the scene while the museum was working to restore normal operations.

theregister.com3 min read

NEW GHOSTGPT CHATBOT FACILITATES MALWARE CREATION AND PHISHING

Researchers from Abnormal Security have reported on a new malicious chatbot that is being sold via Telegram. They believe that GhostGPT is a wrapper around a jailbroken version of ChatGPT or other open-source LLM. This post highlights the risk that these tools bring by giving low-skilled hackers more powerful tools.

infosecurity-magazine.com2 min read