The primary destination for leading cybersecurity news and discourse. Powered by @TechstrongGroup.
Joined April 2010
- Tweets41.6K
- Following1.6K
- Followers7K
- Likes3.4K
Security Boulevard retweeted
Every October, people pay good money to walk through haunted houses and have someone with a plastic chainsaw jump out at them, and IT professionals can get a comparable experience by opening an email that begins, "We made a small change and now nothing works."
The 13 scariest things in IT include a server nobody owns but nobody can turn off, an AI agent given administrator privileges because broad permissions made deployment easier, and a backup that has never met a restore test.
Read the full list, ranked from mildly unsettling to pager-waking: buff.ly/ujhLI3w
#IT #Cybersecurity #AIAgents #DevOps #ITOperations
Security Boulevard retweeted
Cursor disclosed a security incident involving malicious versions of Anthropic's Claude Code extension that spent roughly 11 hours in its marketplace on Oct. 29, attempting to install a persistent backdoor that inserted an attacker-controlled SSH key labeled svc-telemetry, established persistence through a scheduled task on Windows, a LaunchAgent on macOS and a cron entry on Linux, and collected the names of up to 40 environment variables tied to AWS and Google Cloud credentials.
Programming errors prevented some of the most damaging functions from working on many machines, but a failed remote connection does not necessarily mean a computer is free of malicious software.
Here's what affected developers should check and change: buff.ly/L9vuw4V
#Cybersecurity #SupplyChainSecurity #Cursor #ClaudeCode
Security Boulevard retweeted
Enterprise AI controls, open-source AI bans and smarter body cameras. @mvizard @ashimmy @jswartz @TeriRnNY @elizabethsafran @jpmorgenthal #AI #OpenSource #Governance #BodyCameras #DevOps nitter.cf/i/broadcasts/1YxNrblaM…
Security Boulevard retweeted
A poisoned 1.1 GB pnpm-store cache, written from a low-trust pull request and restored by the legitimate release workflow, was enough to publish 84 malicious TanStack packages in six minutes without anyone touching the publish workflow or stealing an npm token.
That makes caching a security boundary, and if state is writable in a low-trust context and reachable from a privileged one, it is a flaw.
Read the full piece on how cache poisoning crosses the trust boundary and how to break the chain without giving up the speed: buff.ly/nJVZtL4
#DevSecOps #CICD #SupplyChainSecurity #AppSec #InfoSec
Security Boulevard retweeted
"Keep a human in the loop" is the most trusted safety answer in institutional AI, and at agent speed and agent scale it becomes one of two things, a bottleneck or a fiction.
The real question was never whether humans belong in the loop but where their authority should sit, and the answer is calibrated placement at the reasoning layer, where aligned interpretations proceed in milliseconds and only the divergent few reach a reviewer with real evidence and real authority.
Read the full argument on why the blanket version of human oversight is obsolete: buff.ly/AftYi2F
#AI #AIAgents #AIGovernance #HumanInTheLoop #EnterpriseAI
Security Boulevard retweeted
Agreed. I’d log the intended recipient, channel and message before the independent egress block rejects the attempt. That lets the eval score attempted persuasion without involving a maintainer. Would you test whether the agent switches channels after denial?
CrowdStrike says a China-linked group called VAPOR PANDA used the DeepSeek AI model and the ARTEX hacking framework to run an 11-day campaign against more than 200 organizations in at least 14 countries, gaining access to nine and targeting government agencies, pharmaceutical companies and chipmakers.
Rather than human operators driving each stage, specialized AI agents shared information and executed attacks simultaneously, with one intrusion processing roughly 1.4 billion input tokens, and the new challenge is that attackers can now test numerous routes into networks without adding personnel.
Read the full article on what CrowdStrike found and how defenders should respond: buff.ly/cBU2APh
#Cybersecurity #AIAgents #DeepSeek #CrowdStrike
Hackers stole an estimated $6.3 billion from Binance in about 20 minutes by chaining a suspected compromised DevOps laptop to the exchange's hot wallet signing infrastructure, with TRM Labs attributing the theft across 10 blockchains to North Korean group TraderTraitor with high confidence.
The lesson experts keep returning to is containment, since sophisticated attackers do not need to defeat every control, only one trusted pathway that lets them operate undetected.
Read the full article on the breach points under investigation and what defenders should take from it: buff.ly/pxllsdb
#Cybersecurity #Binance #Crypto #InfoSec #TraderTraitor
NIST is giving federal agencies and technology suppliers a timetable to phase out the public-key cryptography future quantum computers could break, with its new Internal Report 8547 discouraging RSA, ECDSA, EdDSA and Diffie-Hellman schemes after 2030 and no longer permitting them under NIST standards after 2035.
The guidance folds in federal deadlines requiring agencies to move high-value assets to post-quantum key establishment by Dec. 31, 2030 and digital signatures by Dec. 31, 2031, and it urges hardware vendors to start integrating post-quantum support now given lengthy development and manufacturing timelines.
Read the full article on the deadlines, the recommended migration path and why TLS and IKE come first: buff.ly/bGzL7U5
#Cybersecurity #PostQuantum #NIST #Cryptography #InfoSec
A DHS data leak exposed records on 41,000 workers, including undercover personnel and sensitive surveillance operations, raising concerns about targeted cyberattacks and employee safety.
Read more: buff.ly/nhdd9rx
#Cybersecurity #DataBreach #DHS #DataSecurity #ThreatIntelligence
Cisco CTO Tom Gillis told ONUG attendees that a Shield capability already applies user-space controls to the Linux kernel via eBPF and will soon extend to the applications running on it, letting teams mitigate a vulnerability without patching production.
That matters because as AI models make it possible to understand the relationship between 50 to 100 million lines of code, adversaries can craft targeted exploits in minutes, and no team will apply every patch in time.
Read the full article on how Cisco is buying defenders time: buff.ly/NWT4Ghm
#Cybersecurity #eBPF #Cisco #AppSec #ZeroTrust #InfoSec
AI has made bug discovery almost a commodity, and some vendors have responded with a tantalizing pitch: fully autonomous remediation, where the AI not only finds the bug but ships the fix.
Only 26% of more than 6,000 automated patches tested by researchers actually fixed the target vulnerability without introducing new problems, and unsupervised agents have already deleted a live production database and its backups, which is why human review isn't a "bottleneck" any more than quality assurance is.
Read the full argument for automating the right steps while keeping engineers where judgment matters most: buff.ly/TLODxH4
#Cybersecurity #AppSec #AIAgents #VulnerabilityManagement #DevSecOps
Security Boulevard retweeted
AWS launched Continuum for code vulnerabilities, using AI to detect, prioritize and generate fixes for security flaws. Human approval remains required before patches are applied.
Read more: buff.ly/WdpfC6R
#AWS #DevSecOps #Cybersecurity #AI #ApplicationSecurity
Security Boulevard retweeted
The window for exploiting a vulnerability has shrunk from a year to a day, yet most teams still patch application by application, rebuilding and retesting every stack even when a single shared dependency is what's vulnerable.
Standardizing on common base images and runtimes, reusing patched components instead of redistributing entire image layers, and pointing AI agents at buildpacks and orchestrators can turn a hundred rebuilds into one update.
Read the full piece on why the update process is often the weakest link in vulnerability management: buff.ly/0MlGTal
#Cybersecurity #AppSec #DevSecOps #VulnerabilityManagement #Containers #Kubernetes #AgenticAI
Security Boulevard retweeted
If you are a founder of a small or medium business, this is food for thought.
Only 2% of businesses report having implemented firm-wide cyber resilience, yet resilience goes beyond continuity planning because corrections and downtime still leave customers unable to access certain functions when a crisis hits.
Zero-trust networks verify every access request before approval, unifying security policy and applying least privilege so an organization can keep delivering minimal viable service levels instead of scrambling after an incident.
Read the full piece on the steps to implementing zero trust effectively: buff.ly/aJqj1CV
#Cybersecurity #ZeroTrust #OperationalResilience #InfoSec #SASE
Online scams doubled to roughly 6.4 million attacks in the first quarter of 2026 and passed traditional phishing for the first time, as AI lets criminals generate dynamic templates, multilingual lures and condition-aware evasion at scale while credential theft gets harder.
The fight against fraud is no longer an isolated financial function but a cross-functional cybersecurity imperative.
Read the full report on how AI industrialized fraud: buff.ly/WjiX72D
#Cybersecurity #Fraud #Phishing #AI #Akamai
Security Boulevard retweeted
What do you think about putting all of a tech media company’s content behind a reg wall? Check out my poll in LinkedIn and let me know what you think? linkedin.com/posts/alanshime…
Google is introducing Multi-Party Compute, a security platform built for banks that need stronger protection for blockchain transactions and digital asset custody, distributing the credentials that authorize transfers across multiple computing environments instead of keeping a full signing key on one system.
Stolen private keys accounted for 43.8% of cryptocurrency theft in 2024, and Alphabet has already deployed the technology to protect its own assets on Ethereum and Solana, with a commercial launch expected in early 2027.
Read the full article on how the platform works and where it can still be fooled: buff.ly/ngEP9NE
#Cybersecurity #Google #Crypto #DigitalAssets #Blockchain
The Federal Election Commission's electronic filing system has been offline since Monday evening after an attacker tried to breach the agency's systems, shutting down web-facing services and blocking political committees from submitting required campaign finance reports one week before the Nov. 3 midterms.
Voting will not be affected, but with super PACs and dark money groups averaging more than $50 million a day in spending, voters may lack timely information about who is financing candidates and political advertising in the final days.
Read the full article on the outage and what it means for disclosure: buff.ly/TNXVpCU
#Cybersecurity #FEC #ElectionSecurity #CampaignFinance #InfoSec
