Brief #129: 7 ChatGPT Vulnerabilities, New OWASP 2025 Top 10, 2 Million Jobs Myth
Samsung Galaxy zero-day exploited by LANDFALL spyware for surveillance across Middle East. Social engineering attacks surge 1,450% in H1 2025 with breakout times under 60 minutes.
OWASP's 2025 Top 10 introduces Software Supply Chain Failures as a new #3 category, while social engineering attacks jumped 1,450% in the first half of 2025
ChatGPT vulnerabilities are enabling attackers to steal private data from hundreds of millions of users through 0-click attacks and memory poisoning techniques
The "two million cybersecurity jobs" narrative gets debunked with real data showing only 514,000 actual openings, explaining why qualified professionals struggle to find work
RoboShadow is a cybersecurity platform that delivers daily vulnerability scans and automated fixes, helping businesses stay secure and compliant. With powerful integrations and continuous monitoring, it makes managing cyber risks simple and proactive.
OWASP has published the Top 10:2025 release candidate featuring two new categories and one consolidation, with Broken Access Control maintaining its #1 position while Security Misconfiguration jumps from #5 to #2 based on data from 2.8 million tested applications.
The new Software Supply Chain Failures category expands beyond vulnerable components to encompass broader ecosystem compromises including build systems and distribution infrastructure, ranking #3 despite limited testing data but showing the highest average exploit and impact scores from CVEs.
A tenth new category called Mishandling of Exceptional Conditions addresses improper error handling and logical errors, while the methodology now analyzes 589 CWEs (up from 400 in 2021) grouped into categories focusing on root causes rather than symptoms.
The Mandos Brief
Get this kind of analysis every week. Subscribe to the Mandos Brief.
Researchers uncovered LANDFALL, a previously unknown Android spyware family targeting Samsung Galaxy devices through CVE-2025-21042, a zero-day vulnerability in Samsung's image processing library that was actively exploited in the wild before being patched in April 2025.
The spyware was delivered via malicious DNG image files likely sent through WhatsApp, enabling comprehensive surveillance capabilities including microphone recording, location tracking, and collection of photos, contacts, and call logs on targeted devices in the Middle East.
LANDFALL shares infrastructure and tradecraft patterns with commercial spyware operations, indicating possible links to private-sector offensive actors (PSOAs), with similarities observed to Stealth Falcon activity and references to "Bridge Head" terminology commonly used by commercial spyware vendors.
A new cybercriminal umbrella brand called Scattered LAPSUS$ Hunters (SLH) emerged in August 2025, appropriating reputational assets from Scattered Spider, ShinyHunters, and LAPSUS$ groups while operating through a federated model rather than a formal merger of the original entities.
The group operates primarily through Telegram channels that have been recreated at least 16 times following takedowns, offering Extortion-as-a-Service capabilities and targeting cloud-first environments including SaaS providers, CRMs, and database systems for data theft and extortion.
SLH demonstrates advanced technical capabilities including AI-automated vishing, exploit development targeting CVE-2025-61882 (Oracle E-Business Suite) and CVE-2025-31324 (SAP NetWeaver), with evidence suggesting fewer than five core operators control approximately 30 public personas through sophisticated sockpuppetry.
The report analyzes security team structures across three company sizes: Fortune 500 organizations with 50+ security FTE generating over $7 billion annually, large enterprises ($1-10 billion revenue with 15-49 FTE), and midsize companies ($50 million-$1 billion revenue with 3-15 FTE).
Compensation levels vary significantly by organization size, with Fortune 500 functional department leaders earning an average of $307,000 annually – 25% higher than large organizations and 40% higher than midsize companies for equivalent roles.
Fortune 500 security organizations show mature structures with 31% having dedicated deputy CISO roles, 95% of CISOs engaging regularly with boards, and over 50% utilizing MSSPs primarily for threat detection and response services.
Europe-based victims constitute nearly 22% of entities named on ransomware dedicated leak sites, making it the second most targeted region after North America with over 2,100 victims named since January 2024. The UK, Germany, Italy, France and Spain were the most targeted nations, with manufacturing, professional services, and technology being the primary sectors impacted.
Vishing campaigns are emerging as a significant threat vector for European entities, with CrowdStrike observing nearly 1,000 vishing-related incidents globally during the reporting period. SCATTERED SPIDER notably targeted UK retail entities in 2025, demonstrating sophisticated social engineering techniques including attempted close-access operations at corporate headquarters.
Russia-aligned threat actors continue intensive cyber operations supporting the Ukraine conflict, with GRU-nexus adversaries like FANCY BEAR and VOODOO BEAR conducting intelligence collection and destructive operations against Ukrainian critical infrastructure while also targeting European allies providing support to Ukraine.
ClickFix social engineering campaigns experienced a 1,450% increase in the first half of 2025, making social engineering the second most common initial access method at 39% after business email compromise. These attacks use fake CAPTCHA pages to trick users into executing malicious PowerShell commands that deploy Remote Access Trojans.
Threat actors significantly reduced breakout times to under 60 minutes on average, with some cases showing lateral movement in less than 15 minutes. Attackers primarily use Remote Desktop Protocol to access an average of five hosts per incident while leveraging RMM tools and tunneling techniques for persistence.
Lumma Stealer re-emerged as the most consistently observed infostealer, while Remote Access Trojans including NetSupport RAT, Remcos, and AsyncRAT gained traction among threat groups like TA569 and APT33. The overall incident rate jumped from 6% of customers in H2 2024 to 17% in H1 2025.
A DFIR analyst with several years of experience is questioning their career path due to constant stress and pressure to keep up with naturally talented colleagues, feeling they must grind hard just to remain average in the field.
The professional expresses concern about sacrificing their best years for work stability rather than pursuing personal interests like surfing, training, or building a business, leading them to consider treating work as merely a financial engine.
Community responses reveal widespread burnout across cybersecurity roles, with many suggesting alternative paths within the field such as GRC, CTI, or security management positions that offer better work-life balance while utilizing existing skills.
Experienced professionals with 5+ years are finding that remote positions have become extremely competitive, with many companies now requiring local or hybrid arrangements and immediately discarding applications from candidates not already in the area.
Job seekers report success rates improving when focusing on networking and local opportunities, with several professionals landing positions through referrals, internal connections, and in-person relationship building rather than cold applications.
The market shows inflated requirements where entry-level positions now demand 7+ years of experience, while professionals with decades of experience face rejection for being "not senior enough," indicating a challenging landscape for both new and experienced candidates.
The ISC2 Workforce Study's 4.76 million gap represents modeled staffing needs versus current workforce, not actual funded job openings, while CyberSeek shows only 514,000 U.S. listings with 18-22% being ghost postings that are never filled.
Qualified candidates face systemic barriers including ATS filters that auto-reject viable applicants, experience inflation demanding 2-3+ years for "entry-level" roles, and phantom job postings used for pipeline building rather than active hiring.
Organizations can improve hiring by labeling postings honestly, separating required from preferred qualifications, implementing skill-based screening over rigid filters, and offering structured apprenticeships with remote work flexibility.
Check Point Research demonstrated two complementary approaches to using generative AI for malware analysis: live MCP integration with IDA Pro and x64dbg, and an offline pipeline that exports IDA databases to ChatGPT's cloud environment. The offline approach proved particularly effective, allowing analysis without persistent local sessions while enabling reproducible results and easy collaboration across research teams.
The research team successfully analyzed XLoader 8.0, one of the most challenging malware families that uses multiple encryption layers, obfuscated API calls, and runtime-only code decryption. Using GPT-5 in thinking mode, they reduced analysis time from days to hours, automatically identifying three distinct function decryption schemes, implementing working decryptors, and extracting 64 hidden C2 domains from encrypted configuration data.
While AI significantly accelerated reverse engineering tasks like triage, deobfuscation, and script generation, human expertise remained essential for the most sophisticated protections. XLoader's scattered key derivation logic and multi-layer function encryption still required manual analysis and targeted adjustments, demonstrating that AI serves as a force multiplier rather than a complete replacement for security analysts.
Security researchers conducted the first empirical comparative analysis of three major AI agent communication protocols - Google's A2A, CORAL, and ACP - using a 14-point vulnerability taxonomy covering authentication, authorization, integrity, confidentiality, and availability domains.
The study revealed critical implementation flaws despite strong architectural designs, with CORAL showing robust transport-layer validation but suffering from authentication and authorization failures at its SSE gateway, while ACP's optional JWS enforcement creates high-impact integrity and confidentiality vulnerabilities.
Researchers identified a security dichotomy between architectural strength and implementation weakness, recommending a hybrid approach that combines CORAL's integrated architecture with ACP's mandatory per-message integrity guarantees to address vulnerabilities like prompt injection, data exfiltration, and replay attacks.
Tenable researchers identified seven vulnerabilities in ChatGPT, including novel indirect prompt injection techniques that can exfiltrate user memories and chat history without user knowledge. The flaws affect both ChatGPT 4o and the latest GPT-5 model, bypassing OpenAI's safety mechanisms designed to protect users from such attacks.
The vulnerabilities enable multiple attack vectors including 0-click attacks through web search results, comment-based injections on trusted sites, and persistent memory poisoning. Attackers can exploit users simply by asking ChatGPT innocent questions, as malicious prompts hidden in indexed websites get processed during web searches.
Researchers demonstrated complete attack chains including phishing, data exfiltration using Bing redirect links to bypass url_safe protections, and persistent memory injection that continues stealing data across sessions. The techniques affect hundreds of millions of ChatGPT users and could enable large-scale data theft through strategically placed malicious content.
Israeli cybersecurity startup OneLayer secured $28 million in Series A funding led by Maor Investments, bringing their total funding to over $43 million since emerging from stealth in 2022. The company competes against established players like Fortinet, Juniper Networks, and Trend Micro.
OneLayer currently serves approximately 30 private network customers across North America, Latin America, Europe, and the Middle East, including utilities, manufacturing facilities, mining operations, seaports, and airports. Notable clients include Southern Linc and Evergy in the U.S. market.
The company's platform addresses security blind spots in private 5G/LTE networks by providing unified visibility of all connected assets, supporting anywhere from hundreds to 2 million devices depending on network scale. Their solution bridges the gap between legacy IT systems using IP/MAC addresses and cellular networks using IMSI/IMEI identifiers.
Cloud security company Zscaler has acquired SplxAI to strengthen its zero-trust capabilities with AI asset discovery, automated red teaming, and governance features that protect AI applications throughout their entire lifecycle from development to deployment.
Splx brings over 5,000 purpose-built attack simulations designed to uncover vulnerabilities in AI systems, along with real-time remediation suggestions to fix security issues before they become exploitable by threat actors.
The acquisition follows Zscaler's purchase of Red Canary and expanded partnership with CrowdStrike, positioning the company to provide comprehensive protection across AI and data lifecycles as enterprises invest heavily in AI infrastructure spending forecast to reach $375 billion globally in 2025.
The funding round was led by Goldman Sachs Alternatives' Growth Equity fund with participation from CapitalG, Alphabet's venture arm, positioning the company for its planned IPO in 2026 or 2027.
Armis provides cyber exposure management and security platforms to help businesses secure internet-connected devices, achieving over $300 million in annual recurring revenue in August 2024.
CEO Yevgeny Dibrov stated the company aims to reach $1 billion in ARR before going public, citing increased demand for their IoT security and device management solutions.
A cloud native application protection platform that provides unified visibility, risk assessment, and remediation capabilities across multi-cloud and hybrid environments.
TensorOpera AI is a platform that provides tools and services for developing, deploying, and scaling generative AI applications across various domains.
CalypsoAI is a platform that provides centralized security, observability, and control for deploying and scaling large language models and generative AI across an enterprise.
If you found this newsletter useful, I'd really appreciate if you could forward it to your community and share your feedback below!
How do you like Mandos Brief?
TerribleBadOkayGoodExcellent
Thank you!
Your input helps me make Mandos even better.
P.S. I am working with select B2B companies on the exact challenges covered above. Calendar link here if you'd like to chat.
Talk to you in the next one.
Best,
Nikoloz
MCP Access
Query this dataset from your own AI.
Point Claude, ChatGPT, or Cursor at the Mandos MCP server and ask it about any cybersecurity vendor, product, funding round, or acquisition. Self-serve, pay only for what you pull.
