TLDR InfoSec

TLDR INFORMATION SECURITY 2025-06-04

6 sections18 entrieslinks blockcomplete parse

Parse notes

  • Subscription/account footer truncated before parsefooter_truncated

01GENERAL

Sponsored

THREATSPIKE RED: UNLIMITED PENTESTING, ONE FIXED PRICE

Pentesting prices are a CFO's worst nightmare. $50K for a web app test. Another $30K if you want APIs included. Scope creep? That'll be extra. Need a retest after fixes? Start the procurement process again.

threatspike.comSponsor

02ATTACKS & VULNERABILITIES

ANDROID MALWARE CROCODILUS ADDS FAKE CONTACTS TO SPOOF TRUSTED CALLERS

The Crocodilus Android malware can now add fake contacts to trick users into thinking calls are from trusted people or companies. It has spread worldwide and added new tools to avoid detection and steal better data. Experts warn users to only download apps from trusted sources and use Google Play Protect.

bleepingcomputer.com2 min read

03STRATEGIES & TACTICS

HIGH LEVERAGE SECURITY DECISIONS

A security engineer outlines five crucial decisions for startups to minimize security risks: using an identity provider like Google Workspace or Okta ($6-15/month), implementing hardware security keys for phishing-resistant MFA ($50-75/user), and adopting Infrastructure as Code tools like Terraform for version control. The guide also suggests leveraging managed cloud services such as AWS Fargate and RDS over self-managed VMs, and deploying Mobile Device Management to secure employee endpoints from the outset. The author highlights that significant security improvements stem from operational excellence and strategic tool choices rather than additional security processes or analysts.

0xda.de10 min read

DEVOPS TOOLS TARGETED FOR CRYPTOJACKING

Wiz researchers discovered a cryptojacking campaign called "JINX-0132" targeting misconfigured DevOps servers, including Nomad, Consul, Docker, and Gitea. Attackers exploit default settings and vulnerabilities to deploy XMRig miners, avoiding custom malware by using GitHub repositories. The campaign affects 25% of cloud environments, with 5% exposed and 30% misconfigured.

wiz.io10 min read

HOW TO DEPLOY AI SAFELY

Microsoft's Deputy CISO for AI outlines key principles for safe AI deployment: understand potential failures and create plans, analyze entire systems, including humans, continuously assess risks, and maintain written safety plans. AI errors are intrinsic—treat AI like "well-intentioned but inexperienced new hires." Emphasize extensive testing over coding time and careful information presentation across human-AI boundaries.

microsoft.com5 min read

04LAUNCHES & TOOLS

Sponsored

4 BROKEN PARTS OF CYBERSECURITY

RSA was filled with promises of AI and shift-left. But talk to teams on the ground and it's a different story, filled with unaddressed risks, bolted-on tools, and security that slows down development. Here are the four ways the current approach is failing - and how develop-first security fix them [11] on the Symbiotic blog [11].

symbioticsec.aiSponsor

DELEGATIONS

The Delegations tool enables management of all types of Kerberos delegations (unconstrained, constrained, and resource-based) in Active Directory.

github.comGitHub repository

SBOMMV

Sbommv is a tool for transferring SBOMs between systems. It is designed to fetch SBOMs from input sources, translate and validate them, enrich metadata, and push them to output destinations.

github.comGitHub repository

PASSWORD PUSHER

Securely share sensitive information with automatic expiration and deletion after a set number of views or duration, with full audit logs. Track who, what, and when.

github.comGitHub repository

05MISCELLANEOUS

MICROSOFT, CROWDSTRIKE LEAD EFFORT TO MAP THREAT ACTOR NAMES

Microsoft and CrowdStrike are leading an effort to map different names for the same cyber threat groups. This helps security teams better understand and track threats without needing a single standard name. Other companies like Google and Palo Alto Networks are joining to make threat info clearer and more useful.

securityweek.com2 min read

META AND YANDEX ARE DE-ANONYMIZING ANDROID USERS' WEB BROWSING IDENTIFIERS

Meta and Yandex exploit browser protocols to bypass Android security, linking web browsing to app identities. Their tracking code sends identifiers from websites to Facebook, Instagram, and Yandex apps via localhost ports, de-anonymizing users. Google is investigating this violation of privacy principles. Both companies have confirmed they have paused the practice.

arstechnica.com9 min read