Brief #105: npm Package Attacks, Cybersecurity Offshoring Trend, AWS Strands Agents

Happy Sunday!

The FBI's latest IC3 report caught my eye, showing a 33% jump in cybercrime losses to $16.6 billion is sobering - especially seeing investment fraud taking the top spot. Human psychology is still one of the major target points.

In this week's brief:

• 60 malicious npm packages discovered exfiltrating network data through Discord webhooks
• Third-party involvement in data breaches has doubled to 30% according to Verizon's DBIR
• Companies are increasingly offshoring cybersecurity roles while expecting remaining staff to "do more with less"

What's your take - are we approaching a breaking point where security teams simply can't keep up with both traditional threats and new AI-powered attacks using current staffing models? Let me know in comments or by replying to this email.

Industry News

Threat Actor Claims 1.2 Billion Facebook Records For Sale With Suspicious Inconsistencies

• Threat actor "ByteBreaker" claims to have scraped 1.2 billion Facebook records by abusing an API, offering sensitive data including full names, email addresses, phone numbers, and location information.

• Multiple inconsistencies in the listing raise doubts about its legitimacy - including mismatched record counts (claims of 1.2B records but sample data showing "Total Rows are 200 million") and conflicting Telegram contact information.

• Despite questionable claims, API abuse remains a serious threat vector that has affected numerous platforms including LinkedIn, Chess.com, and Trello - requiring organizations to continuously monitor for unusual activity.

60 Malicious npm Packages Leak Network and Host Data in Active Campaign

• Socket's Threat Research Team discovered 60 malicious npm packages across three accounts that collect and exfiltrate network data through a Discord webhook, with over 3,000 downloads targeting developer environments.

• The malicious packages use install-time scripts to gather internal/external IPs, hostnames, DNS servers, and directory paths while using sandbox-evasion techniques to specifically target real production environments.

• All packages contain identical reconnaissance code designed to map organization networks for future intrusions, with the campaign remaining active as packages are still available on the npm registry.

SEO Poisoning Attack Redirects Employee Paychecks Through Hijacked Routers

• Attackers used SEO poisoning to create fake login pages targeting mobile devices, stealing employee credentials to access payroll portals and redirect paychecks to attacker-controlled accounts.

• The campaign utilized compromised home routers and residential IP addresses as proxies, making detection difficult while allowing attackers to bypass security measures that would normally flag suspicious login locations.

• The attackers leveraged Pusher service for real-time notification of compromised credentials, enabling immediate reuse before passwords could be changed or suspicious activity detected.

Leadership Insights

FBI's IC3 Reports Record $16.6 Billion in Cybercrime Losses for 2024

• The FBI's Internet Crime Complaint Center (IC3) received 859,532 complaints in 2024, with a record $16.6 billion in reported losses – a 33% increase from 2023, with investment fraud being the top category ($6.57 billion).

• Ransomware complaints increased 9% from 2023, remaining the most pervasive threat to critical infrastructure, with IC3 identifying 67 new variants. The Recovery Asset Team successfully froze $561.6 million through the Financial Fraud Kill Chain process.

• Adults over age 60 were the most targeted demographic, submitting the highest number of complaints (147,127) and suffering the greatest financial losses ($4.8 billion), while cryptocurrency was involved in $9.32 billion of reported losses.

Netwrix Research: 77% of Organizations Now Operate in Hybrid IT Environments

• Netwrix survey of 2,150 IT professionals reveals 60% of organizations already leverage AI tools in their infrastructure, with 37% reporting that AI-driven threats have forced them to adjust their security approach.

• Cloud security incidents are increasingly identity-driven, with user/admin account compromise affecting 46% of organizations in 2025 compared to only 16% in 2020, while targeted attacks on premises rose from 19% in 2023 to 28% in 2025.

• Financial damage from security incidents increased dramatically, with 75% of organizations reporting monetary impact (up from 60% in 2024) and organizations estimating damages of $200,000+ nearly doubled to 13%, highlighting growing ransomware concerns.

Third Parties Involved in 30% of Data Breaches According to Verizon DBIR

• Third-party involvement in breaches doubled from 15% to 30% in the past year, with notable incidents involving stolen credentials accessing platforms like Snowflake where the median time to remediate leaked secrets in GitHub repositories was 94 days.

• Ransomware grew 37% from last year, now present in 44% of all breaches, with small businesses disproportionately affected (88% of SMB breaches vs 39% in larger organizations), though median ransom payments decreased to $115,000.

• Exploitation of vulnerabilities as an initial access vector increased by 34%, reaching 20% and approaching stolen credential usage rates, with edge devices and VPNs becoming primary targets (22% of exploits, up from 3% last year).

Discover my collection of industry reports, guides and cheat sheets in Cyber Strategy OS

Career Development

U.S. Cybersecurity Job Market Faces Offshoring and AI Pressures

• Companies are increasingly offshoring cybersecurity roles to Latin America, Eastern Europe, and South Asia, with some organizations reporting 70% of their security staff now located overseas.

• The "do more with less" mentality has led to significant reductions in force, with some professionals handling work that should require entire teams, while relying heavily on automation to manage workloads.

• Executive hype around AI capabilities is driving additional layoffs, though industry professionals predict this may reverse as organizations discover AI cannot perform at expected levels.

New Security Engineer Faces Limited Mentorship and Training in Contract Position

• A new contractor with three months experience as a security engineer expressed frustration about minimal training, being told to "read documentation," "email account owners," or "copy similar tickets" rather than receiving proper guidance on tools like CrowdStrike, Splunk, and Palo Alto.

• Industry professionals responded that self-sufficiency is critical in cybersecurity careers, with most recommending the engineer leverage available resources to learn independently rather than waiting for formal training that may never materialize.

• Multiple commenters advised sticking with the position at least six months, as this timeframe typically represents when employees become fully comfortable in their roles, while using the access to enterprise tools as an opportunity for self-directed learning.

2025 Cyber Security Salary Survey Shows Shifting Market Dynamics and Priorities

• Cyber security recruitment faced challenges in 2024 due to economic uncertainty, with candidate supply outweighing employer demand, particularly at senior levels, though 57% of employers plan to hire additional staff in 2025.

• Remote working has emerged as a critical factor in recruitment, with 74% of professionals stating they would consider changing jobs if not permitted flexible arrangements, and 40% valuing it more than annual bonuses.

• The AI governance field is experiencing growth, with demand rising for professionals with relevant skillsets as organizations navigate data privacy and security challenges posed by artificial intelligence implementation.

AI & Security

Anthropic Develops Method to Monitor AI Value Expressions in Real-World Conversations

• Anthropic's researchers have created a privacy-preserving system that analyzes how their AI assistant Claude expresses values during real-world user interactions, examining 700,000 anonymized conversations to build the first large-scale empirical taxonomy of AI values.

• The study found Claude generally follows its intended "helpful, honest, harmless" training, but researchers detected some instances where values like "dominance" and "amorality" appeared, likely resulting from user jailbreak attempts to bypass AI safety guardrails.

• This monitoring approach offers a new method for AI alignment evaluation that could detect problematic behaviors emerging only in real-world settings, though it cannot be used pre-deployment and requires substantial conversation data to implement effectively.

Meta Releases LlamaFirewall: An Open Source Guardrail System for AI Agents

• Meta has released LlamaFirewall, an open-source security framework designed as a final layer of defense against risks associated with AI agents by addressing prompt injection, agent misalignment, and code vulnerabilities.

• The system includes three main guardrails: PromptGuard 2 (a universal jailbreak detector), Agent Alignment Checks (a chain-of-thought auditor), and CodeShield (an online static analysis engine to prevent generation of insecure code).

• Already in production at Meta, LlamaFirewall includes customizable scanners that allow developers with basic regex or prompting skills to quickly update security guardrails for their AI agents.

AWS Releases Strands Agents, An Open Source AI Agents SDK

• AWS has released Strands Agents, an open source SDK that enables building and running AI agents in just a few lines of code, with a model-driven approach that leverages LLMs' reasoning capabilities.

• The SDK is already used in production by multiple AWS teams including Amazon Q Developer, AWS Glue, and VPC Reachability Analyzer, and is designed to scale from simple to complex agent use cases.

• Strands supports multiple LLM providers including Amazon Bedrock, Anthropic, Ollama, Meta, and others, with companies like Accenture, Anthropic, Langfuse, mem0.ai, Meta, PwC, Ragas.io, and Tavily joining as contributors.

Market Updates

StackHawk Raises $12 Million To Address AI-Powered Development Security Challenges

• The shift-left API security platform secured funding from Sapphire and Costanoa Ventures to help AppSec teams cope with the increasing pace of AI-driven development, bringing total funding to $47.3 million.

• Developers now deploy code 70% faster using AI, but 30% contains security vulnerabilities, creating a widening productivity gap that hackers exploit as security teams struggle with the 100:1 developer-to-security professional ratio.

• StackHawk will focus on developing solutions that help identify and remediate critical API security issues during development, particularly for data-sensitive industries like healthcare and fintech.

Rhino Federated Computing Raises $15M Series A For Secure AI Collaboration Platform

• Boston-based Rhino has secured $15M in Series A funding led by AlleyCorp to expand its federated AI platform, bringing total funding to $30M since its 2020 founding.

• The company's technology enables organizations in regulated industries to collaborate on AI model development without sharing sensitive data, addressing critical data privacy concerns.

• Founded by Dr. Ittai Dayan, former AI leader at Mass General Brigham, and Yuval Baror, the company employs 20 people with half based in Israel, positioning itself at the intersection of AI development and regulatory compliance.

BreachRx Secures $15M to Scale Incident Response Platform

• Incident response startup BreachRx has raised $15 million in Series A funding led by Ballistic Ventures to expand its go-to-market and engineering teams.

• The platform offers a centralized workspace for security, legal, compliance, and communications teams, automating response plans and defining clear roles and responsibilities during an incident.

• BreachRx recently launched Rex AI, a generative artificial intelligence engine that streamlines incident response by providing real-time recommendations and automating administrative tasks.

Tools

Blackpanda Managed Security Services

Blackpanda Managed Security Services offers a comprehensive cybersecurity solution through their IR-1 subscription model that combines incident response, vulnerability scanning, and cyber insurance access.

Cyver Core

Cyver Core is a pentest reporting and management platform designed to streamline the entire penetration testing workflow. The platform supports the complete pentest journey from scoping to remediation through five key phases:

Symbiotic Security

Symbiotic Security is an AI-powered code security solution that integrates directly into the IDE to detect, remediate, and educate developers about security vulnerabilities in real-time.

If you found this newsletter useful, I'd really appreciate if you could forward it to your community and share your feedback below!

For more frequent cybersecurity, leadership and AI updates, follow me on LinkedIn, BlueSky and Mastodon.

Best, 
Nikoloz