A critical vulnerability in LMCache, open-source software that speeds up large language model (LLM) servers such as vLLM, lets an attacker run code on the cache server without logging in, and no fixed version is available. The flaw is in LMCache's multiprocess mode, where the cache runs as a standalone server that LLM workers reach over the ZeroMQ messaging library. A single network
Cybersecurity News and Vulnerability Aggregator
Cybersecurity news aggregator
Updated: 10:20 pm
J/K Navigate / Search S Save V Open Link
— or
treemd <(curl -sL https://allsec.sh/md) (as Markdown) Top Cybersecurity Stories Today
AWS penetration testing is scoped by accounts, identities, and exposed workloads, not IP ranges. AWS lets customers test a defined list of services without prior approval, but it prohibits flooding, DNS attacks, and takeovers, and it requires approval for command-and-control and simulated events. Budget follows two drivers: account and IAM complexity for the cloud control plane, and the number of internet-facing applications for workload testing. The post AWS Penetration Testing for Enterprises: What to Scope and How to Budget appeared first on Synack .
Latest
Wednesday, October 7
At a recent event for security firm NordVPN, NBA superstar Shaquille O’Neal revealed that he got hacked—and warned the public about the need to “have control of their own information.”
It's 11pm, you're parsing through mountains of logs. There's yet another tool someone remembered, and they have the logs exported. The time zone is in PT or ET, or something that isn't UTC. You are tired and questioning your life choices because this thing is not in UTC and now you need to convert the time zone in your head. I made a simple [Clock app](https://github.com/rdmershon/PowerShell-Clock-App) to make my life easier. I'm sharing it in the hopes it helps others. It uses no external dependancies. It uses no third-party libraries. It just uses native win forms and PowerShell. I like a lot of DFIR tools, but you can't always bring them into some environments. This is just a basic PowerShell script so it should be able to be run in most environments.
Attackers compromised three country-code top-level domains (ccTLDs) and obtained unauthorized HTTPS certificates for several Google domains, Google said on October 6. Google's own systems were not breached, but any domain ending in .gh (Ghana), .sl (Sierra Leone) or .as (American Samoa) was put at risk. With such a certificate, an attacker could pose as the real site over an encrypted
Hi guys, I send out a weekly newsletter with the latest cybersecurity vendor reports and research, and thought you might find it useful, so sharing it here. All the reports and research below were published between September 28th - October 4th. You can get the below into your inbox every week if you want: [https://www.cybersecstats.com/cybersecstatsnewsletter/](https://www.cybersecstats.com/cybersecstatsnewsletter/) # Big Picture Reports **Relentless Defense (Cisco)** There are “Relentless Defenders,” and then there’s everyone else, which is what this report claims. It also claims to explore what the best-prepared organizations do differently so you can be like them. **Key stats:** * Only 8% of organizations surveyed qualify as “Relentless Defenders”, the report’s top-performing group for coverage, response speed and organizational alignment. * 40% of security teams say they spend more time manually collecting and correlating data than dealing with threats. * 39% say critical insights are missed because the data sits somewhere they cannot reach. *Read the full report* [*here*](https://www.cybersecstats.com/r/06f2b551?m=50f43416-1146-4a3d-a1e1-5afc95e09a39)*.* **2027 Global Digital Trust Insights (PwC)** A useful large-scale report from PWC to compare notes about where cyber budgets are going, which threats others feel least prepared for, and how much they trust AI to act on its own. **Key stats:** * 84% of security and finance leaders expect their cyber budget to increase, up six percentage points from last year. * Half of security leaders identify attacks targeting AI systems as one of their biggest preparedness gaps. * 55% rank the reliability and maturity of AI technology among their top three barriers to increasing agent autonomy. *Read the full report* [*here*](https://www.cybersecstats.com/r/b53e4634?m=50f43416-1146-4a3d-a1e1-5afc95e09a39)*.* **2026 Digital Defense Report (Microsoft)** Another big report. Microsoft’s annual look at the threat landscape, including credential theft, data theft, and how quickly exposed cloud workloads attract attacks. **Key stats:** * Exposed cloud workloads are attacked an average of 5.3 hours after exposure. * 63% of intrusions involve data theft. * Between 89% and 95% of email phishing attachments lead to credential theft efforts. *Read the full report* [*here*](https://www.cybersecstats.com/r/3bf0fa6a?m=50f43416-1146-4a3d-a1e1-5afc95e09a39)*.* # AI Security **2026 Identity Security Report: The AI Enforcement Gap (Delinea)** What’s the gap between organizations’ AI access policies and their ability to enforce them? **Key stats:** * 99.7% of organizations have a formal policy governing which data AI tools and agents can access. * 87% of IT leaders say an AI tool or agent accessed sensitive data beyond its intended scope in the past year. * 55% of organizations take a full day or longer to detect when an AI agent steps outside its scope. *Read the full report* [*here*](https://www.cybersecstats.com/r/0ef89734?m=50f43416-1146-4a3d-a1e1-5afc95e09a39)*.* # Inside the SOC **The New SOC Career Ladder: How AI Is Reshaping the SecOps Workforce (Swimlane)** AI seems to be making SOC work more satisfying. Though some worry it’s leaving them with fewer opportunities to build their skills. **Key stats:** * 88% of security operations professionals and leaders say AI makes their work more satisfying. * 24% say AI limits their skill development. * Among those who say AI limits their skill development, 91% still report higher job satisfaction. *Read the full report* [*here*](https://www.cybersecstats.com/r/a161d98b?m=50f43416-1146-4a3d-a1e1-5afc95e09a39)*.* # Vulnerability Management **H2 2026 Cyber Hygiene Index (Detectify)** Comparing how long organizations in different regions take to fix critical and high-severity vulnerabilities. **Key stats:** * 90% of open critical and high-severity vulnerabilities have been exposed for more than 90 days across the organizations analyzed. * That figure reaches 97% in the Nordics. * In the UK, it is 92%, compared with 86% in the US. *Read the full report* [*here*](https://www.cybersecstats.com/r/d1b541ef?m=50f43416-1146-4a3d-a1e1-5afc95e09a39)*.* **Vulnerability Discovery and Exploitation Trends in the AI Era (Google Threat Intelligence Group)** AI is busy finding security holes. Meanwhile, researchers are finding holes in AI. **Key stats:** * Monthly vulnerability disclosures more than doubled, from 5,045 in January 2026 to 10,740 in August. * From January to August 2026, 141 distinct vulnerabilities were disclosed and exploited, already exceeding the 127 recorded across all of 2025. * Only 0.23% of vulnerabilities disclosed in 2026 (about one in 431) were observed in active exploitation. *Read the full analysis* [*here*](https://www.cybersecstats.com/r/358ce6c3?m=50f43416-1146-4a3d-a1e1-5afc95e09a39)*.* # Social Engineering **The Deepfake Readiness Index 2026 (Pindrop)** US enterprises’ readiness for deepfake attacks, the impact of incidents and the attention these threats get in the boardroom. **Key stats:** * 74% of US enterprise security leaders have encountered or suspect a deepfake attack in the past year. * Only 10% report having purpose-built tools to address deepfake threats. * 75% of security leaders say it will take a company leader being personally fooled or impersonated before deepfakes become a genuine boardroom priority. *Read the full report* [*here*](https://www.cybersecstats.com/r/fa2033e1?m=50f43416-1146-4a3d-a1e1-5afc95e09a39)*.* # Skills and Training **2026 SkillBit Micro-Training Survey Report (SkillBit)** Finding the right security people is only the start. **Key stats:** * 57% of cybersecurity leaders say new hires take six months to reach full productivity. * Nearly 64% of organizations consider three months an acceptable time-to-value for new cybersecurity hires. * 70% report having few or no roles available to candidates with less than two years’ experience. *Read the full report* [*here*](https://www.cybersecstats.com/r/82ef5bf9?m=50f43416-1146-4a3d-a1e1-5afc95e09a39)*.* # Industry-Specific **2026 Data and Identity Security Report: Healthcare Findings (Netwrix)** Who’s got access to sensitive healthcare data? With AI adding more people and tools to the mix, keeping track is getting harder. **Key stats:** * 79% of healthcare organizations say their non-human identities are not fully governed. * 77% cannot immediately determine who has access to a specific piece of sensitive data. * 31% experienced unauthorized identities accessing sensitive data in the past year, compared with 24% in other industries. *Read the full report* [*here*](https://www.cybersecstats.com/r/300454ea?m=50f43416-1146-4a3d-a1e1-5afc95e09a39)*.* # Committed to the Mission: The State of the DIB with CMMC in Flux (Redspin) Where defense contractors are investing and what they’re prioritizing to meet cybersecurity requirements. **Key stats:** * 75% of defense contractors say achieving CMMC Level 2 certification provides value beyond contract eligibility. * 78.2% of organizations are continuing towards CMMC certification or are already Level 2 certified by a third party. * 21.9% are delaying certification or significantly slowing their implementation and certification efforts. *Read the full report* [*here*](https://www.cybersecstats.com/r/6dbf0bb8?m=50f43416-1146-4a3d-a1e1-5afc95e09a39)*.*
Cybersecurity researchers have disclosed details of a long-running npm supply chain malware campaign that pushes information stealers and remote access trojans (RAT) to compromised hosts. The campaign has been codenamed MALFEX by CloudSEK and Checkmarx. The activity is assessed to be the work of a lone threat actor who appears to have published 12 packages since August 2023, eight of which have
I started a blog series to provide free insights into appsec. It’s mainly to breakdown what application security is all about and it’s mainly targeted towards beginners and startups, so take it as you will. I want to teach every one interested in appsec my perspective on it from my experience in big tech. So far: know your app → lock down who gets in → follow untrusted input → catch broken business logic → build a threat model. Part 7 puts it all to the test. Literally. Prove which problems are real before you spend time fixing them. Please reach out if you have any questions or would like for me to write on a topic that you’d want to learn more about.
Hundreds of new repositories poisoned with malicious, secret extracting workflows. GitHub held most malicious workflows runs for approval, limiting the impact. Some victims found to also be affected by seemingly unrelated cryptominer attacks, hinting toward a pool of compromised credentials being used by multiple threat actor groups.
My team wrote a technical walkthrough of the distributed alerting workflow we use internally. It uses an agent to investigate GuardDuty findings, identify owners, and handle responses in Slack. Slack helped popularize distributed alerting by sending alerts directly to the people involved and asking whether they recognized the activity. Dropbox and Brex described similar approaches, but most teams we’ve spoken with struggled to scale beyond user-centric alerts where someone’s identity was already in the payload. Our security engineer ran into this at his previous company. A suspicious login was straightforward to route, but an alert about a host or infrastructure change could take hours or days of work with detection engineers before they could reliably identify someone to ask. We walk through a concrete Kubernetes example where the alert only names a service account. Finding a person means following the evidence through workload metadata, code ownership, deployment history, and identity logs. The tutorial shows how we turn that investigation method into reusable skills so an agent can work through those steps for each finding. We cover: * The history of distributed alerting and why deterministic workflows struggle to scale * The estimated effort of building owner lookups across GuardDuty finding types * How the agent uses skills, drilldown queries, and context from inventory, code, tickets, and logs * The full flow from investigation to a Slack conversation and the owner’s confirmation * Permissions, tool restrictions, observability, and keeping context current What interested us most was extending this to host and infrastructure alerts that had been impractical to automate with individual lookup rules. The walkthrough includes the setup and examples so you can see how it works. We know many folks were are divided by the use of agents in security automation. Hopefully this article on distributed alerting presents an example of something that wasn't possible / too time consuming to work before agents + context / skills driven automation vs fully determined paths. [https://www.tracecat.com/blog/agentic-security-alerts](https://www.tracecat.com/blog/agentic-security-alerts)
SonicWall has released hotfixes for four flaws in its SMA1000 appliances, the gateways that give remote workers access to a company's network and applications. The most serious could allow an attacker without a login to send requests through the appliance and reach internal functions. SonicWall rates it 10.0 on the CVSS scale and says it has no evidence that any of the four flaws is being
A critical vulnerability in LMCache, open-source software that speeds up large language model (LLM) servers such as vLLM, lets an attacker run code on the cache server without logging in, and no fixed version is available. The flaw is in LMCache's multiprocess mode, where the cache runs as a standalone server that LLM workers reach over the ZeroMQ messaging library. A single network
The US government’s Tradewinds initiative has made it easier to throw millions of dollars at “nontraditional” defense contractors, including OpenAI, Anthropic, and Google.
Cybersecurity researchers are calling attention to a new malware family that has been observed targeting exposed artificial intelligence (AI) and large language model (LLM) infrastructure with an aim to deploy cryptocurrency miners and further expand the scale of the botnet. The financially motivated campaign, dubbed Canto Incognito, has been found to install cryptocurrency miners, including
AWS penetration testing is scoped by accounts, identities, and exposed workloads, not IP ranges. AWS lets customers test a defined list of services without prior approval, but it prohibits flooding, DNS attacks, and takeovers, and it requires approval for command-and-control and simulated events. Budget follows two drivers: account and IAM complexity for the cloud control plane, and the number of internet-facing applications for workload testing. The post AWS Penetration Testing for Enterprises: What to Scope and How to Budget appeared first on Synack .
Log extractions (such as a compromised `.claude.json` file) show the stealer successfully exfiltrating raw `primaryApiKey` values and detailed OAuth account data tied to Anthropic/Claude accounts. By grabbing these CLI tokens, attackers are bypassing traditional web logins entirely, gaining direct, programmatic access to premium AI models, organizational workspaces, and potentially sensitive source code passing through these tools.
\> ML4 is one of the world's strongest AI models for cybersecurity. On the Artificial Analysis Cyber Index, an independent evaluation of how well AI models find and fix security flaws in real software, it ranks among the top five models globally and leads open-weight models developed outside China by a wide margin. Their blog talks a lot about the cyber capabilities of their new model. Its quite interesting.
According to the new FBI/USSS advisory, FortiBleed actors no longer just add persistence accounts. In some cases they also delete or reset the original admin accounts (T1531), which locks the owner out of their own FortiGate. That changes the usual playbook. If you assume a password reset gets you back to a clean state, it won't help when you can't log in at all. A few things worth checking: → Do you have out-of-band admin recovery for your edge devices? → Have you audited REST API keys? They survive password resets. → Is SSH left open on the firewall? How are others handling recovery for edge devices? Full breakdown: [https://hubs.la/Q04zrkbN0](https://hubs.la/Q04zrkbN0) Free FortiBleed checker: [https://hubs.la/Q04zrk9\_0](https://hubs.la/Q04zrk9_0) Advisory: [https://www.ic3.gov/CSA/2026/261006.pdf](https://www.ic3.gov/CSA/2026/261006.pdf)
The 2026 findings are not just a year-over-year shift. They mark the latest point in a five-year arc where resilience, AI governance, human risk, and board scrutiny are converging inside the systems where work actually happens. For years, the enterprise cybersecurity story has been told as a straight line of escalation: more attacks, more data loss, more pressure, and more urgency. That
The U.S. Federal Bureau of Investigation (FBI) and Secret Service (USSS) on Tuesday warned that the FortiBleed credential harvesting campaign remains an active threat aimed at internet-facing Fortinet FortiGate firewalls and secure socket layer (SSL) virtual private network (VPN) gateways. "The campaign exploits reused or leaked credentials and legacy SHA-256 password storage, enabling threat
Threat actors have begun to exploit a newly disclosed critical security flaw impacting Atlassian Data Center products that could allow access to sensitive files under certain conditions. The arbitrary file access flaw, tracked as CVE-2026-21589 (CVSS score: 9.3) affects multiple products, including Bitbucket Data Center, Confluence Data Center, Jira Service Management Data Center, Jira Software
If you’re evaluating an agentic pentesting solution right now, you’ve probably heard the same pitch more than once: point it at a target, and it discovers, validates, and exploits attack paths autonomously, the way a real attacker would. That promise is worth taking seriously. It’s also worth pressure testing, and three questions do the heavy lifting. What can the assessment actually
Cybercrime Detective Explains Cybersecurity Jargon in 60-Second Videos for Cybersecurity Awareness Month
COPENHAGEN, Denmark, 7 October 2026 – Heimdal today officially launches the “Cyber in 60” weekly video series in which Adam Pilton, a former cybercrime detective and now Cybersecurity Advisor at Heimdal, breaks down one cybersecurity term or concept in under 60 seconds. This series solves an old tech to human translation problem. The people who […] The post Cybercrime Detective Explains Cybersecurity Jargon in 60-Second Videos for Cybersecurity Awareness Month appeared first on Heimdal Security Blog .
Anthropic on Tuesday said it's expanding a program that allows vetted cybersecurity professionals to test its advanced artificial intelligence (AI) models with reduced safeguards and blocking classifiers, as the company claimed its Project Glasswing initiative uncovered at least 129,000 verified software vulnerabilities between April and July 2026. The company said it also found an additional
The Computer Emergency Response Team of Ukraine (CERT-UA) has identified more than 100 compromised websites that have been injected with malicious JavaScript to serve an information-stealing malware called LunexStealer (aka Psychedelic Stealer). The activity, which was observed by the agency in September 2026, has been attributed to a threat cluster dubbed UAC-0277. It did not disclose who the
Dark web roundup: 10M French residential records, 19M SMTP creds, U.S. manufacturer VPN/RDP access, and a Struts exploit kit (unverified)
Sharing a few recent dark web listings that came up in monitoring. None of these are confirmed yet, so take them as claims for now, not confirmed breaches. → France: a seller claims 10M+ residential records → IUT Paris Seine: 6.8GB of data plus about 30M log entries → U.S. manufacturer: VPN and RDP access up for auction → SMTP dump: about 19M credentials → Apache Struts: an exploit tool sold together with access to servers already compromised The Struts listing stands out to me. Selling the tool together with working access points more to an access broker than to someone trying to sell a PoC. The 19M SMTP creds are the likeliest to show up again soon in phishing and credential stuffing. Has anyone seen overlap with the SMTP dump in their own environment, or seen recent Struts exploitation in the wild? Full breakdown:https://hubs.la/Q04z7HkK0
OpenAI came to Australia to apologise. It will leave without answering these key questions | Toby Walsh
We don’t accept drugs killing a few people because they cure most patients. Or a few planes crashing because most land safely. AI should be no different OpenAI’s CEO Sam Altman was asked to appear before the Joint Select Committee on Artificial Intelligence in Australia and explain how and why their agents have been hacking into multiple government websites . Altman didn’t turn up but sent Jason Kwon, his chief strategy officer, instead to answer the committee’s many tricky questions on its first day of public hearings on Tuesday. The committee is conducting a comprehensive inquiry into the economic, societal, regulatory, and national security implications of AI for Australia. It will report back to parliament at the end of November. This is an unusually fast pace, but then it’s an unusually fast paced technology. Continue reading...
Tuesday, October 6
Cybersecurity researchers have disclosed details of a "human-operated phishing platform" that impersonates advertising products for artificial intelligence (AI) chatbots like Google Gemini, Anthropic Claude, OpenAI ChatGPT, Perplexity, Meta Muse, and Manus. The products, which claim to offer campaign optimization, spend audits, and business-account connections, are designed with one goal in
Linux backdoors targeting telecom and network appliances in South Korea and Taiwan have been disguising their traffic as email services and seemingly legitimate processes to blend in and evade detection. Threat actors are known to name their malicious software after a legitimate operating system component or a process as a defense evasion measure. By borrowing the name of a real binary, it may
Shares in fashion company fall nearly 10% after its app systems are accessed by ‘unidentified third party’ Asos is investigating unauthorised access to its app system after shoppers received a notification claiming hackers had “fully compromised” its data. The online fashion retailer said basic personal information including name and contact details might have been accessed by an unidentified third party but it did not believe payment card records or passwords had been compromised. Continue reading...
You Won’t Hear About These, Even In Myths (Atlassian Jira, Confluence (and more) Pre-Auth Arbitrary File Read CVE-2026-21589)
Welcome back to yet another episode of "security was taken seriously". Being who we are (and constantly being exposed to what we see…), we recognize we have been doomed to eternal damnation as we keep on watching security best practices crumble behind “secure by design” public statements. And in the times we live in, where anyone with a prompt window in front of them can say "reproduce the vulnerability, make no mistakes", so are you. In Greek mythology, Atlas was punished by the gods for misbehaving. Reality is unfair, and all we get is Atlassian punishing the rest of us for running their systems on-prem. And so they did, on October 5th, in a security advisory . Sometimes we regret that CVSS scores go as high as 10, because when a vu
I made this javascript deobfuscator for https://github.com/javascript-obfuscator/javascript-obfuscator
I wanted to share a project I’ve been working on that I’m super excited about: Project RedTeam: Contract Offensive There’s a free Demo that provides a tutorial and lets you play a few contracts (no time limit, play as much as you want). Some players are already pulling some serious hours in the demo! At its core, this is a game about using MITRE ATT&CK adversarial techniques against procedurally generated networks. It's delivered in a gameplay loop that plays a lot like Balatro and other card based Roguelike games. In Project RedTeam, you need to earn money to pay off debts after every contract within a run. Earn money by completing objectives, side bounties, or executing exfiltration/ransom against targets- the choice on how to be profitable is always yours. It's a challenging but fun and fast paced take on network-intrusion cybersecurity concepts. It's entertaining in a deliberately gamified way. A goal of this project was to create a hacking game that is realistic enough to keep it meaningful as a tool to teach intrusion concepts and stages to anyone- but not be overcomplicated and slow-paced like most hacking games. I've put a lot of thought into the design and dynamics of how to capture the core-loop of network intrusion and turn it into a game that's approachable. The design direction of this project is an outcome of having over a decade of training and experience in cybersecurity. Feel free to AMA! I'm happy to answer any questions about the game and/or development process :) Give the game a Wishlist on Steam or share this post if it's something you support and want to see further development on. Project background: This was implemented over the past 3 months using a modern development workflow (yes, modern AI tools make this possible- I'm not hiding that fact!). That being said, this is by far the most complex software project I've built as a solo developer and it was not an easy or simple development task. There's a Steam Community with a Dev Blog for this game that provides more history/progress updates on the project. Mods: This will be my only and last post here, since it is promotional. I just wanted to share this since there’s been very positive interest from similar subreddits.
The Wikimedia Foundation, which hosts Wikipedia, has confirmed that it has discovered activity by rogue OpenAI agents on its platforms, including unsuccessful efforts to compromise Etherpad, a public note-taking tool, and edit Wikipedia pages. "The unauthorized bot activities included edits to our wikis, some unsuccessful attempts to exploit a public note-taking tool we host, and heavy traffic,
In 2024, MCP (Model Context Protocol) set out to become the USB-C of AI: one standard for connecting models, agents, and IDEs to tools and data. The protocol delivered. Thousands of developers built servers, and enterprises plugged them into agent workflows. The ecosystem around it fell short. Earlier this year, our team at OX Security, traced critical vulnerabilities in Anthropic's MCP
Google has stopped accepting product vulnerability reports through its bug bounty program for its open-source software. The change, in effect since October 1, means researchers can no longer submit security flaws in the code of projects such as Go, Angular, and Protocol Buffers there for a reward. Reports about supply chain compromises are still accepted, and reports filed before October 1 are
Project Zero often works with software vendors to remediate the vulnerabilities we report and provide broader guidance on making software more secure. Some vendors express concern about potential scenarios in which they are unable to fix vulnerabilities that are causing immediate user harm, due to limitations in their patch delivery systems. Since Project Zero encounters a wide array of systems designed to protect users in the case of exceptional exploitation scenarios, both through vendor discussions and security reviews, we want to share what we’ve learned. This post provides an overview of systems in use by large vendors that allow them to remediate small volumes of vulnerabilities much faster than their typical update process. Our goal is to provide a reference for vendors seeking to implement or enhance the capabilities of such systems, and to encourage vendors to consider how they would fix an urgent vulnerability before they receive one.
The U.S. Federal Bureau of Investigation (FBI) has removed an Accenture contractor for their alleged role in a ShinyHunters-breach that led to the theft of personal details of thousands of bureau employees. That's according to a report from Reuters, citing two sources familiar with the matter. "To date, our review has determined that the incident occurred as the result of a security failure
Unauthorized parties have gained access to the names, addresses, and personal identification numbers of about 8.8 million people, living and dead, in Denmark's national population register, the country's digitalization ministry said on October 5. They used a private Danish company's lawful right to look up records in the Central Person Register (CPR). The ministry has told people never to
In this post, we share LLM-specific validation patterns that attackers use to test exposed AWS credentials for Amazon Bedrock access.
Monday, October 5
A year since launching Synack’s integration with Qualys, the partnership now includes broader platform support, more AI-led testing, and a shared presence at Qualys ROCon Americas 2026. The post Turn Scanner Findings into Validated Risk with Synack and Qualys appeared first on Synack .
To reduce the amount of noise from questions, we have disabled self-posts in favor of a unified questions thread every week. Feel free to ask any question about reverse engineering here. If your question is about how to use a specific tool, or is specific to some particular target, you will have better luck on the [Reverse Engineering StackExchange](http://reverseengineering.stackexchange.com/). See also /r/AskReverseEngineering.
I wrote up my investigation into @goodjavascript/dotenv@1.0.0, including how to recover its entry-point file after the npm tarball and jsDelivr file URLs returned 404. The useful detail: Software Heritage’s latest snapshot contained only the security placeholder. An older snapshot retained the original release and its 840-byte index.js. Its SHA-256 matched the digest still available in jsDelivr’s file manifest. Static inspection showed a timer scheduled at module load that collects host information and can execute JavaScript supplied in a server response. The package had no installation scripts, and its exported config() function was empty. The article includes the package-to-archive discovery steps, dated evidence, an annotated code excerpt and a Python verifier that retrieves and hashes the file without executing it. It also links my analysis contribution to the existing OSV advisory. [https://cgsec.dev/research/dotenv-recovery/](https://cgsec.dev/research/dotenv-recovery/) This concerns the scoped @goodjavascript/dotenv package, not the unscoped dotenv package. Have you used other archives or retained metadata sources to recover removed package evidence?
We analyzed RBAC bindings across over 65,000 Kubernetes clusters to find dangerous permissions granted to system:anonymous, system:unauthenticated, and system:authenticated.
Sunday, October 4
Presently sponsored by: Where are your AI agents? Origin's sensor finds every install on your fleet, grouped by owner, including the ones your MDM never sees. I'm in Denmark! Well, just, I'm now at Copenhagen airport ready to begin the long trek home, with the final event at GOTO now done and going just perfectly. This week, there are two ShinyHunters arrests in the news: Pepijn in the Netherlands and then Saif in Jordon . It's an inevitable outcome, of course, and as I say this week, it was also the most likely one. Time will tell how many more join their ranks, but the seriousness of the crimes, the length of time they were perpetrated over, and the motivations behind them will certainly see substantial custodial sentences. In other news, this week I'm properly introducing a new sponsor for the blog: Origin . One of our next AI frontiers is understanding what agents have actually done (and we've all seen news of where they're been a bit too, well, "creative" in executing their tasks), and Origin's solution gives you visiblity into just that. Check them out, and a big thanks to them for their ongoing support.
David Robinson joins other insiders in urging industry to take more care over rapidly developing technology A safety leader at OpenAI has quit the company, warning that its culture was broken and that AI firms were not “being nearly careful enough” about developing the technology. David Robinson, who led the writing of safety reports that accompanied the ChatGPT developer’s product releases, explained his resignation in an essay headlined: “I quit OpenAI because its culture is broken.” Continue reading...
Saturday, October 3
Pre-installed C2 loader on cheap Android projectors - deploys proxy/ad-fraud botnets, can run arbitrary code
Hi everyone, I recently bought one of those cheap Android projectors (Nonete HY260Pro, Allwinner H713) and noticed some suspicious network activity. Being curious, I decided to set up a lab, intercept the traffic, and dig into the firmware. I ended up uncovering a factory-installed malware ecosystem: a disguised dropper (StoreOS), a hidden second stage (SilentSDK) and a plugin loader that talks to a C2 server in China (api.pixelpioneerss.com) and deploys up to 5 botnet/fraud plugins. Key findings of my analysis: * Four-stage infection chain: StoreOS → SilentSDK → PluginManager → final plugins. Payloads are hidden with a "Byte-Reversal" trick, XOR encryption and build-fingerprint spoofing. * The plugins currently enroll the device in residential proxy networks (XFJ/net2fast, a UDP proxy node, indicators consistent with the Vo1d botnet) and run ad/click fraud, partly geo-fenced server-side. * The loader executes downloaded code with system privileges, so the operators can push any payload at any time. I only observed proxy and ad-fraud plugins, but the capability for remote code execution is there by design. * The device also ships with open root backdoors and sends device identifiers (MAC, serial, Android ID) to servers in China. * An independent researcher found identical infrastructure on a Magcubic HY300 Pro+, which suggests the problem is the H713 firmware base and not one brand. I could only verify my own device. This is my first independent technical report and deep dive into malware research. I've documented the full kill chain, decrypted the obfuscated strings, listed IOCs and mitigations (DNS blocklist, ADB disable commands), and written scripts to repair the malformed payloads for analysis. Full Report: [https://github.com/Kavan00/Android-Projector-C2-Malware](https://github.com/Kavan00/Android-Projector-C2-Malware) I'd love to get your opinion on the report, especially from owners of other H713 devices who can compare. Looking forward to your feedback!
Millions have downloaded Meta’s AI agent Muse. But getting it to do your bidding comes with privacy costs.
Internxt is a post-quantum secure encrypted cloud storage provider which is open-source and has passed multiple independent audits. I reviewed their code and found that post-quantum security should have been the least of their problems. Clicking a link in your browser could trigger remote code execution on the desktop app or leak your long-term encryption keys to an attacker-chosen URL. Their cryptographic architecture stands on shaky grounds with public keys never being verified, in some cases man-in-the-middled by design, a flat key hierarchy and a KDF with just 3 iterations of MD5. We need PQC and we need it now, but adding a (self-rolled) PQC hybrid on top of a weak protocol does not make it more secure.
Friday, October 2
DHS agents not only tracked and intimidated people observing ICE activity in Maine, but stored information about them in a database run on Palantir software, newly unsealed court filings say.
When analysing firmware attack surfaces, image decoders built into bootloaders get less scrutiny than cryptographically verified OS kernels. If image parsing happens before signature verification, any memory corruption in the parser breaks the secure boot trust model. researchers analysed U-Boot's video subsystem (drivers/video/video\_bmp.c) and identified an unbounded write in the RLE8 bitmap decoder (video\_display\_rle8\_bitmap()) that leads to a **pre-authentication Secure Boot bypass**. **Root Cause and Vulnerability Mechanics** When **U-Boot** displays a boot logo or splash screen, it parses a BMP image loaded from local storage (SPI flash, MMC/eMMC, USB, or SD card). **Unbounded framebuffer write:** During RLE8 decompression, `video_display_rle8_bitmap()` decodes run-length encoded streams directly into the active framebuffer without validating stream bounds against the frame boundary or allocated buffer size. **Pre-authentication execution window:** In many embedded target configurations, the boot splash screen is rendered immediately on startup, before U-Boot calls Android Verified Boot (AVB) or FIT image signature verification routines. **Storage disparity:** The kernel image and rootfs are signed, but **splash images are frequently stored in unsigned, user-writable partitions or external media**. An attacker who writes a crafted RLE8 BMP to the boot storage can trigger an out-of-bounds write past the framebuffer during early boot, corrupting adjacent bootloader data structures, function pointers, or verification flags in memory. This hijacks the execution flow before signature checking completes.
**TL;DR.** [SConnect](https://chromewebstore.google.com/detail/sconnect/mjhbkkaddmmnkghdnnmkjcgpphnopnfk) \- 1M+ users, an extension middleware+native host for authentication with eIDs, 3SKeys and other hardware signing tokens had a drive-by RCE which enabled any site or iframe a user saw to silently download and execute a dll due to a poor hand-rolled implementation of RSA-2048 token validation, enabling a use of uninitialized memory validation bypass which enabled "plugins" (DLLs) to be loaded. v2.16.0.0 of the extension and native host is vulnerable. [CVE-2026-18397](https://nvd.nist.gov/vuln/detail/cve-2026-18397). CVSS 9.4.
API Connections allow anyone to fully compromise any other connection worldwide, giving full access to the connected backend.
Multihashing is one of those cryptographic tasks that’s easy not to think about too much. This is unfortunate, because multihashing is a common stumbling point when cryptographers try to use hashes. As part of our goal to “fix software, not bugs,” Trail of Bits is introducing SequenceHash and its sister function SequenceMAC , a pair of related hash constructions that bring secure multihashing to developers using hash functions other than Keccak. We hope SequenceHash and SequenceMAC will help cryptographers avoid attacks that take advantage of ambiguous input encodings. The specification is open source, and is now a part of the Community Cryptography Specification Project (C2SP). SequenceHash and SequenceMAC behave similarly to NIST’s TupleHash , but have the advantage of not being tied to a single hash function. They also don’t require developers to implement fiddly computations that aren’t byte-aligned. Instead, SequenceHash and SequenceMAC work out of the box with nearly any secure cryptographic hash function you care to use, including SHA256/384/512, BLAKE, and RIPEMD. SequenceMAC supports keys 32 bytes or longer (up to the ridiculous limit of ${2}^{128}-1$ bytes). (It’s worth noting: SequenceHash and SequenceMAC rely on the security of the underlying hash for their own security. SequenceHash and SequenceMAC can’t magically make MD4 or SHA0 secure again. For the purposes of this document, it’s assumed that you have chosen a reasonable hash function like SHA256, not CRC32.) To make SequenceHash and SequenceMAC easy to use, we’re releasing
Introduction Pwn2Own is a renowned hacking competition organized by the Zero Day Initiative (ZDI), where security researchers demonstrate previously unknown vulnerabilities in popular software, operating systems, browsers, IoT devices, and other technologies. Having participated in both the 2023 and 2024 editions of Pwn2Own, we decided to take another shot in 2025. This time, our goal was to avoid collisions, where multiple teams discover the same vulnerability during the same event, leading to reduced prize money and fewer Master of Pwn points. This blog post walks through our journey from discovery to full exploitation. We start by exploring the Home Assistant device architecture, then detail how we found a remote code execution vulnerability in an add-on. From there, we show how we leveraged it to pivot to the underlying operating system and achieve root-level access. We conclude with our experience at the Pwn2Own 2025 Cork edition. Target Selection We started by looking at several different targets. Our initial list included the Wyze Cam Pan v3 and the Synology CC400W from the surveillance system category, the Brother MFC-J1010DW from the printer category, the Philips Hue Bridge and the Home Assistant Green from the smart home category. After assessing the various targets, we shifted our focus to the Home Assistant Green due to the progress we had made on that platform. This led us to the discovery of an exploit chain that resulted in an unauthenticated remote code execution vulnerability. Device Overview
While the focus has been on AI agents’ hacking capabilities, a recently patched vulnerability in a ChatGPT app shows that AI software is itself an inviting—and vulnerable—target.
Thursday, October 1
I've been picking through a SCADA-style telemetry capture and I can't explain what I'm seeing. The file has 14 authorisation records, each with its own CRC. All 14 validate. The file header carries a CRC-32 over the whole block, and that one fails. Two of the operator-name fields are zeroed. My reading is that someone blanked those fields, recomputed the per-record CRCs so they'd pass, and never touched the block CRC. But I'd like a sanity check — is there a corruption mode that breaks a block checksum while leaving every record checksum intact? File (8.8MB): www.[st88openocean.github.io/slip-three/MCA-CTL-0314-RAW.slice](https://st88openocean.github.io/slip-three/MCA-CTL-0314-RAW.slice) Published digests for it are here: www.[st88openocean.github.io/slip-three/archive](https://st88openocean.github.io/slip-three/archive)
Traverse barriers and checkpoints in the occupied West Bank—your choices reflect the reality of nearly 3.5 million Palestinians who live there.
Before you build an AI pentesting agent, run it through 5 tests for production readiness. Mark Kuhr breaks down what separates a prototype from a system. The post Build or Buy AI Pentesting? 5 Tests to Judge Production Readiness appeared first on Synack .
A blue-team writeup on detecting a compromised MikroTik from its own config. Seven techniques, each with the collection command, the artifact it leaves behind, and a triage step. Feedback welcome.
Overview An Out-of-bounds Write vulnerability in the InsydeH2O IHISI software used by HP PC BIOS can allow a local attacker with OS kernel privileges to perform arbitrary physical memory writes, including writes to System Management RAM (SMRAM). Because the vulnerable code executes in System Management Mode (SMM), successful exploitation can allow an attacker to modify SMM-protected memory and potentially achieve arbitrary code execution in SMM. Depending on the platform's memory and firmware configuration, the arbitrary physical memory write primitive may also have implications for UEFI firmware update or flash-related operations. Description HP PC BIOS is firmware that initializes and manages hardware components during the boot process and provides runtime services to the operating system and provide the ability to update firmware. The affected system uses InsydeH2O Kernel version 5.5 or earlier. The BIOS includes custom HP SMM handlers that execute in System Management Mode (SMM), a highly privileged CPU execution mode that is isolated from the operating system. CVE-2026-12855 : An Out-of-bounds Write vulnerability in the H19WMIHandlerSmm module (GUID f1946499-571b-44c3-9b9c-cc55210b0c02) allows a local attacker with OS kernel privileges to read or write arbitrary physical memory, including SMRAM, through a Software SMI handler. An attacker with kernel-level privileges can trigger the vulnerable SMM handler by issuing a Software System Management Interrupt (SMI) through I/O port 0xB2 and supplying specially crafted CPU register values. The vulnerable handler does not adequately validate the supplied parameters before performing memory operations, allowing the attacker to influence the physical address and data involved in the operation. Because the affe
Benedict Jones spent years working for McAfee and Sophos. While doing threat research at Sophos, he spotted a problem in mobile threat defence that nobody had actually fixed. He’s now CEO and founder of Trustd Mobile, and the story of how he got there says more about MSP buying decisions than most vendor pitches ever […] The post Innovation wins. Why smaller beats bigger appeared first on Heimdal Security Blog .
In an exclusive interview with WIRED, Paragon Solutions CEO Andrew Boyd reveals the limits of the company’s promise to keep bad actors from abusing its powerful espionage tool.
London, UK, 1 October 2026 – Heimdal, a global cybersecurity provider, today announced a partnership with Elovade, a European IT security distributor with 250 experts across five countries, to bring its unified security platform to managed service providers (MSPs) across the DACH region: Germany, Austria, and Switzerland. The move extends a relationship that began a […] The post Heimdal partners with Elovade to bring unified cybersecurity platform to the DACH region appeared first on Heimdal Security Blog .