Security researchers at Forever Security have shown that one ordinary browser extension could take control of the AI assistants built into five Chromium-based products: Gemini Live in Chrome, Perplexity Comet, Microsoft Edge, Opera Neon and the Claude in Chrome extension. Once the extension was installed, it could access each product's built-in AI with a single click. On Comet, Edge,
Cybersecurity News and Vulnerability Aggregator
Cybersecurity news aggregator
Updated: 5:40 pm
J/K Navigate / Search S Save V Open Link
— or
treemd <(curl -sL https://allsec.sh/md) (as Markdown) Top Cybersecurity Stories Today
A critical security flaw in Issabel Framework, a web-based framework for the open-source unified communications PBX software, has come under active exploitation. The vulnerability in question is CVE-2026-89026 (CVSS v3.1 score: 9.8/CVSS v4.0 score: 9.3), which can allow an unauthenticated remote attacker to execute arbitrary operating system (OS) commands by taking advantage of a hard-coded
VU#212479: Sentry Seer vulnerability allows attacker-controlled input to be executed in a privileged environment
Overview A vulnerability exists in Sentry Seer when the system is configured to automatically hand issues to a coding agent for remediation. Successful exploitation results in arbitrary code execution within the codingâagent environment and access to connected source repositories. This vulnerability is tracked as CVE-2026-90999 . Description Sentry is a software errorâmonitoring and performanceâtracking platform used by developers to detect, diagnose, and understand issues in their applications. It collects telemetry such as exceptions, stack traces, logs, and performance data from applications. Built into Sentry, Seer acts as an automated debugging assistant that converts telemetry into actionable remediation steps and can hand off issues to an integrated coding agent to propose code fixes. Because Sentry front-end projects commonly expose a public DSN (Data Source Name) to allow browsers to submit this telemetry, an attacker can craft and submit malicious events through this public endpoint. When Seer is enabled to automatically pass issues to a coding agent, these attacker-supplied events can traverse multiple trust boundaries. Ultimately, malicious event fields propagate through Seerâs analysis pipeline, transforming into untrusted instructions that the privileged coding agent may execute. The vulnerable workflow is as follows: * Sentry ingests attackerâgenerated exception events submitted through the public DSN. * Seer evaluates whether the event represents an issue eligible for automated remediation. * Seer generates a rootâcause analysis that uses attacker-controlled event fields, including exception messages, stack traces, source context, and breadcrumbs. * The generated analysis is embedded directly into the initial prompt provided to the coding agent.
Parallels Desktop for Mac has a flaw that lets an ordinary local account run code as root, the highest level of access on a Mac, software company JFrog said this week. The attack needs code already running on the machine as a normal user, so it does not work over the network. JFrog says the fix is in Parallels Desktop 27, a version that Intel Macs cannot install. Yuval Moravchick, who leads
A hacker collective pulled down a Flock camera and dumped its data. The files included thousands of videos and logs showing that the device captured 1.6 million images of 50,000 vehicles in 21 days.
Latest
Wednesday, September 16
>"CISA will discontinue the weekly Vulnerability Bulletin at the end of FY26 (September 28, 2026) as part of a broader shift from severityâbased vulnerability management to riskâbased vulnerability prioritization. Newly recorded vulnerabilities remain available on [CVE.org](http://CVE.org), and users should rely on the Known Exploited Vulnerability (KEV) Catalog, CISA Cybersecurity Alerts and Advisories, and vendor security alerts for actionable, riskâbased updates." I can't help but feel that we're at the end of an era. I do appreciate the shifting of resources to risk-based vulnerability discovery and eradication, especially with AI-assisted vulnerability research increasing the number of CVEs by such a large amount. RIP to the weekly bulletin though, I'll miss it.
New [Hunt.io](http://Hunt.io) research extends Bitdefender's SilkParasite report by connecting a SpiceRAT C2 cluster to NodeEdgeRAT and NomadRAT at the infrastructure level. A TLS certificate from TLC, a CA funded by a Chinese state research institute, impersonates Uzbekistan's railway and ties much of the cluster together. The domains spoof named Central Asian government and energy entities, and passive DNS places the activity back to at least mid-2022. All network-side, no compromised hosts or samples. Full report: [https://hunt.io/blog/silkparasite-spicerat-central-asia-infrastructure](https://hunt.io/blog/silkparasite-spicerat-central-asia-infrastructure)
A penetration testing scope defines the assurance an enterprise needs, not just the assets a vendor will touch. Start with the business or compliance objective, then map the system boundary before counting assets. List applications, APIs, infrastructure, roles and integrations separately, specify authenticated and unauthenticated testing, and clarify whether production, staging or both will be [âŠ] The post How to Scope an Enterprise Penetration Test: Free Scope-of-Work Template appeared first on Synack .
Overview A vulnerability in MLflowâs dspy and statsmodels model flavors allows unauthorized pickle deserialization executions despite a safety control. Specifically, the dspy flavor conditionally applies the control based on the model pathâs file extension, and the statsmodels flavor does not apply the control. Description MLflow is an open-source platform for managing machine learning lifecycles, including model packaging, versioning, and deployment. "Flavors" refer to the specialized frameworks through which supported models are stored and loaded. In response to previous vulnerability concerns, MLflow implemented the MLFLOW_ALLOW_PICKLE_DESERIALIZATION safety control to block and disable executing any pickle deserialization and subsequent loads per the userâs choice. When loading models through mlflow.pyfunc.load_model(model) , users must specify a model flavor and path in an MLmodel file. With the dspy flavor, MLflow checks the value of MLFLOW_ALLOW_PICKLE_DESERIALIZATION , and whether the specified model path ends in .pkl . A model path that does not end in .pkl (even if the file is actually a pickle file), will route to a separate branch for pickle deserialization, bypassing the safety control. However, when loading through the statsmodels flavor, there is no check for MLFLOW_ALLOW_PICKLE_DESERIALIZATION at all. Impact Exploitation of this vulnerability allows for arbitrary remote code execution through a malicious pickle-loaded payload, regardless of a user explicitly disallowing pickle serialization, through vulnerable flavor specifications in the MLmodel configuration file. The attack path requires write access to any location from which a use
đ”ïž Hunting SilkParasite / SpiceRAT infrastructure by page hash, TLS cert, and RDP-over-TLS on high ports
Detection-focused writeup on a SpiceRAT C2 cluster connected to Bitdefender's SilkParasite report. Everything here is reproducible from scan data. Hunting hooks: a reused decoy page collapses to one SHA-256 body hash across 13 hosts; a TLC-issued cert (subject azure.uzrailwaystax\[.\]com, SHA-256 in the post) sits on 8 hosts; hardened RDP on unusual high ports (64350, 64330, 65535, 65111) shows up under a tls fingerprint rather than RDP, so filter for that. Two nginx versions (1.29.3 and 1.31.3) recur across the fleet. Full IOC tables (IPs, domains, cert fields, subdomain and passive DNS history) plus the HuntSQL queries are in the post, ready for retro hunts. [https://hunt.io/blog/silkparasite-spicerat-central-asia-infrastructure](https://hunt.io/blog/silkparasite-spicerat-central-asia-infrastructure)
Hi folks, my name is Gal Weizman, I do browser security research. Excited to finally share my recent work, where I managed to hack Chrome, Comet, Edge, Opera and Claude in Chrome using one single browser extension 2 CVEs & $20,000 in bounties đ Hope you like it! [https://forever.security/blog/bragjack-attack-hijacks-every-browser-agent](https://forever.security/blog/bragjack-attack-hijacks-every-browser-agent)
A critical security flaw in Issabel Framework, a web-based framework for the open-source unified communications PBX software, has come under active exploitation. The vulnerability in question is CVE-2026-89026 (CVSS v3.1 score: 9.8/CVSS v4.0 score: 9.3), which can allow an unauthenticated remote attacker to execute arbitrary operating system (OS) commands by taking advantage of a hard-coded
Enterprises in Russia have emerged as the target of three threat activity clusters tracked as NightEagle, Hacking Cat, and Toy Ghouls, according to multiple reports from Kaspersky. The cybersecurity vendor said it has identified attacks mounted by NightEagle (aka APT-Q-95), a threat actor known to be active since at least 2023, that involve new techniques for persistence and lateral movement.
VU#212479: Sentry Seer vulnerability allows attacker-controlled input to be executed in a privileged environment
Overview A vulnerability exists in Sentry Seer when the system is configured to automatically hand issues to a coding agent for remediation. Successful exploitation results in arbitrary code execution within the codingâagent environment and access to connected source repositories. This vulnerability is tracked as CVE-2026-90999 . Description Sentry is a software errorâmonitoring and performanceâtracking platform used by developers to detect, diagnose, and understand issues in their applications. It collects telemetry such as exceptions, stack traces, logs, and performance data from applications. Built into Sentry, Seer acts as an automated debugging assistant that converts telemetry into actionable remediation steps and can hand off issues to an integrated coding agent to propose code fixes. Because Sentry front-end projects commonly expose a public DSN (Data Source Name) to allow browsers to submit this telemetry, an attacker can craft and submit malicious events through this public endpoint. When Seer is enabled to automatically pass issues to a coding agent, these attacker-supplied events can traverse multiple trust boundaries. Ultimately, malicious event fields propagate through Seerâs analysis pipeline, transforming into untrusted instructions that the privileged coding agent may execute. The vulnerable workflow is as follows: * Sentry ingests attackerâgenerated exception events submitted through the public DSN. * Seer evaluates whether the event represents an issue eligible for automated remediation. * Seer generates a rootâcause analysis that uses attacker-controlled event fields, including exception messages, stack traces, source context, and breadcrumbs. * The generated analysis is embedded directly into the initial prompt provided to the coding agent.
Security researchers at Forever Security have shown that one ordinary browser extension could take control of the AI assistants built into five Chromium-based products: Gemini Live in Chrome, Perplexity Comet, Microsoft Edge, Opera Neon and the Claude in Chrome extension. Once the extension was installed, it could access each product's built-in AI with a single click. On Comet, Edge,
Clickfix compromise at the infrastructure level - how KongTuke delivered Clickfix on September 14th without compromising individual websites
Parallels Desktop for Mac has a flaw that lets an ordinary local account run code as root, the highest level of access on a Mac, software company JFrog said this week. The attack needs code already running on the machine as a normal user, so it does not work over the network. JFrog says the fix is in Parallels Desktop 27, a version that Intel Macs cannot install. Yuval Moravchick, who leads
N0va is targeting organizations across North America and Europe with phishing campaigns that impersonate trusted services and abuse legitimate authentication flows. Successful attacks can give threat actors access to valid accounts without relying on obvious malware activity. From there, a single compromised identity can open the door to sensitive data, business systems, and additional cloud
Most of the worldâs governments tell you that you need to be post-quantum prepared by 2030-2035. For example, the US government says critical infrastructure should be key-exchange-prepared by the end of 2030 and authentication-prepared by the end of 2031 (ignore for now that they have the threats exactly backward). Non-critical infrastructure has until the end of 2035. There are a few governments saying critical infrastructure should be prepared by 2029, and many large tech companies, such as Google, Cloudflare, and Microsoft, say they plan to be prepared by 2029. But what happens if Q-Day happens in 2028? I cover this topic in my latest post-quantum books, *Becoming Post-Quantum: Your Guide to Post-Quantum Success* ([https://www.amazon.com/dp/B0HHTBDZ7T](https://www.amazon.com/dp/B0HHTBDZ7T)) and *Becoming Post-Quantum: Your Post-Quantum Project Plan* (Companion Book) ([https://www.amazon.com/dp/B0HJJJ675G](https://www.amazon.com/dp/B0HJJJ675G)). This isnât a rhetorical question. IonQ has recently stated (https://www.ionq.com/news/ionq-publishes-worlds-first-fully-compiled-end-to-end-blueprint-for-breaking-256-bit-elliptic-curve-signatures) that it is realistic that its quantum computers could break traditional asymmetric cryptography used by everyone by 2028. To be clear, a paper is not a working cryptographically-relevant quantum computer. But it is a clear sign that IonQ thinks that it is not out of the realm of possibility. I wrote about this last week: [https://www.linkedin.com/pulse/should-you-post-quantum-prepared-2028-roger-grimes-vo9le](https://www.linkedin.com/pulse/should-you-post-quantum-prepared-2028-roger-grimes-vo9le). Note: It takes a lot more than just stable, logical working qubits to make a cryptographically-relevant quantum computer. Marin Ivezicâs analysis is the best Iâve seen on the subject: [https://postquantum.com/crqc-capability-framework/](https://postquantum.com/crqc-capability-framework/). But getting lots of stable, usable error-corrected qubits is one of the biggest challenges, if not the most challenging part. Most organizations havenât even started an official post-quantum project yet. They are late. They need to start now! ASAP!! But even if your organization has started an official post-quantum project, itâs likely that the estimated âQ-Dayâ (i.e., the day when the first quantum computer breaks traditional asymmetric encryption) that the entire project is running on is probably 2030-2035. What happens if Q-Day happens in 2028 or 2029? Well, youâve got to accelerate your post-quantum project, of course. All of it. It would be a massive immediate refocusing of resources. It would likely result in your organization being less efficient and productive (unless you are a vendor in the post-quantum space). You would have to reach out to your immediate boss, executive management, and the board of directors, if you have one, and tell them what just happened. Then you would gather all the necessary resources for an emergency meeting. You would immediately start your cryptographic inventory. Instead of having a year plus to conduct it on all critical systems, you would start conducting it focused first and best on your top critical systems. Ultimately, you would want your cryptographic inventory finished as fast as possible (e.g., a week, a month, a quarter) on your most critical assets. You would then make emergency decisions regarding what can be reconfigured, what needs to be upgraded, what needs to be replaced, what needs to be decommissioned, and what needs to be physically isolated. You would be making these decisions quickly. You might need to hire extra resources well beyond what you have today to conduct the post-quantum project superfast. Do you have those post-quantum vendors selected today? Do you know any? If not, start figuring out whom to call when. What cryptographic inventory software will you use? None are perfect. None inventory everything. But you need to use one (or more) to help you collect your cryptographic inventory as quickly as you can. You should do your research and make these two selections now, so you'll have them when you need them. \[Hint: You need them now, regardless of Q-Day dates.\] On a practical note, if your adversaries can use quantum computers against you (youâd have to calculate the risk of them actually doing so), you would need to implement monitoring tools, processes, and policies to be able to detect a quantum compromise. Is there a way to detect a successful quantum attack? For key exchange breaks, probably not. Other than you learn that previously protected information seems to be known by unauthorized entities. For authentication breaks, you would look out for fake digital certificates and use, and other related artifacts, such as fraudulent successful logons, fraudulent websites with fraudulent âvalidâ digital certificates. You would need to threat model the various quantum threats against your environment and then create and implement the appropriate protective and detective controls. Â You would need to update your SOC (if you have one) and potentially your incident response policies and procedures. How do you respond to a quantum attack if you actually detect one? How do you respond to stolen data? How do you respond to a fraudulently used âvalidâ digital certificate and its use? In the latter case, you would certainly get the fraudulently used digital certificate revoked, warn stakeholders of the use, and get other related fraudulent assets removed and/or taken down. Ultimately, you need to threat model the various quantum attacks that would be possible against your company and take the appropriate preventative, detective, and incident response controls, like you would against any threat. Only this threat is quantum and new. I would also recommend that you focus on mitigating authentication threats first ([https://www.linkedin.com/pulse/us-government-gets-pqc-threats-exactly-backward-you-should-grimes-sx7ee](https://www.linkedin.com/pulse/us-government-gets-pqc-threats-exactly-backward-you-should-grimes-sx7ee)) over key exchange threats, if you canât do them both at the same time. This goes counter to every other post-quantum expert telling you what to do first. Iâm right. They are wrong. Regardless of whether Q-Day comes early or not, itâs cheaper and better to do it sooner than later. I wrote about that here: [https://www.linkedin.com/pulse/why-become-post-quantum-now-versus-later-roger-grimes-wiboe](https://www.linkedin.com/pulse/why-become-post-quantum-now-versus-later-roger-grimes-wiboe). With this post, I just want you to consider what you and the world are going to look like if Q-Day happens years earlier than most governments are telling you to be prepared for. There is at least one quantum computer vendor that is saying itâs not insane to think it might happen in 2028 or soon thereafter. The possibility is non-zero. You need to make that part of your post-quantum plans. It would be irresponsible not to.
Google has disclosed that a high-severity security flaw in its Pixel Cellular Modem has come under exploitation in the wild. The vulnerability, tracked as CVE-2026-58704 (CVSS score: 8.0), is a privilege escalation flaw. "In Cellular Modem, there is a possible permission bypass due to a logic error in the code," according to a description of the bug in the NIST National Vulnerability Database
A leaked credential shows up in a criminal marketplace, or a vulnerability gets a disclosure advisory, and either one can be weaponized against a real target before most security teams have triaged the alert. Attackers are combining that kind of intelligence with AI-assisted exploitation to accelerate the path from exposure to breach faster than most security programs are built to react.
Acronis has warned that a high-severity security flaw in its Backup plugin for cPanel and Web Host Manager (WHM) deployments has been exploited in the wild. The vulnerability, tracked as CVE-2026-87886 (CVSS score: 7.8), is described as a case of local privilege escalation due to insecure file permissions. It affects the following versions - Acronis Backup plugin for cPanel & WHM (Linux
A hacker collective pulled down a Flock camera and dumped its data. The files included thousands of videos and logs showing that the device captured 1.6 million images of 50,000 vehicles in 21 days.
Help Needed: Missing global perspective/rotation matrix data when converting UMK3 iOS .meshset files to .obj
Two things happened last week, one day apart, and almost nobody connected them. On 11 September, the EU Cyber Resilience Actâs vulnerability reporting obligations came into force. Companies covered by the regulation now have to report actively exploited vulnerabilities within 24 hours and provide a fuller notification within 72. On 12 September, Anthropic CEO Dario [âŠ] The post Slow is a design principle, not a delay appeared first on Heimdal Security Blog .
Widescreen + Online Multiplayer Coop for my Rust WASM Zelda 2 decompilation. Github and playable link coming soon.
A critical security flaw in WSO2 API Manager has come under active exploitation in the wild, according to findings from watchTowr. The vulnerability, tracked as CVE-2026-5430 (CVSS score: 9.8/10.0), is a case of improper verification of a cryptographic signature that could result in account takeover. Hacktron Team has been credited with discovering and reporting the flaw. "JWT authentication
I started a blog series to provide free insights into appsec. Itâs mainly to breakdown what application security is all about and itâs 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. I talked about the semantic problem of authorization last time. In this post, I talk about how you can know what makes an application "hackable" through something called taint analysis with a code example! Hope you all enjoy! Please reach out if you have any questions or would like for me to write on a topic that youâd want to learn more about.
In this post, we walk through different threats to GitHub and how to detect them.
Tuesday, September 15
Point it at a host and it does the boring first hour of an investigation for you. What it does in one run: * Collects the evidence (event logs, registry, more) even while files are locked * Builds a timeline and matches it against thousands of detection rules * Scans for known malicious tools on disk * Checks what's connected, what's running, what looks off * Checks IPs and files against threat intel * Spits out a scored report you can hand to someone No API keys required, has its own web interface. There's a demo mode that fakes an incident so you can try it without needing a real machine to investigate. đ [https://github.com/karim852/Kage-DFIR-toolkit](https://github.com/karim852/Kage-DFIR-toolkit) đ„ïž live demo: [https://demo-kumo-kage.vercel.app/demo/kage/index.html](https://demo-kumo-kage.vercel.app/demo/kage/index.html) Feedback and contributions welcome đ
ResetSpy - Enumerate user accounts and registered authentication methods via the Microsoft Self-Service Password Reset (SSPR) portal
As of August 2026, Microsoft has removed the legacy CAPTCHA from the SSPR flow and replaced it with backend throttling and behaviour-based abuse detection (see [MC1400824](https://mc.merill.net/message/MC1400824)). Microsoft's position is that the backend controls are sufficient to detect and block automated abuse. This change makes it easier to enumerate user accounts and also aids in identifying which additional verification methods are in place (provided they are SSPR enabled). More info in the readme.
Cybersecurity agencies in the United States, the United Kingdom, and the Netherlands have detailed a Windows malware that they say Iran's intelligence service uses to spy on dissidents, journalists, and activists around the world. The malware is controlled via the Telegram messaging app and can copy a target's emails and chat messages, take screenshots, and activate the microphone to record
Cybersecurity researchers have disclosed details of a multi-platform campaign that uses the Message Queueing Telemetry Transport (MQTT) protocol as a communication channel to control Windows and Linux systems. The emerging malware family, codenamed BambooToken, is assessed to be active since at least February 2023 and put to use in attacks targeting organizations across Asia and South America.
Third-party penetration testing means two different things: testing your own organization through an independent provider, and testing or reviewing testing on a supplier, partner, or acquisition target. This guide focuses on the second use case. It works best as a governed risk process, not a document request, built on clear authorization, risk-based scope, defensible evidence, and contractual follow-up. The post Third-Party Penetration Testing: What Security and Procurement Teams Need to Know appeared first on Synack .
As more teams â and now agents â build applications on Cloudflare's Developer Platform, having the right access controls is crucial to allow you to ship safely. After all, the last thing you want is for an agent to make a change in production, just because it was granted more access than it needs. Now, you can give a teammate or agent access to a specific Worker, so that they can only make changes to that application and no other resources in your account. Moreover, weâre giving you four new roles, so you can limit exactly what they can do: The new roles are available today, for all customers. You can assign them to a specific user, so when they log into the dashboard, they will only see the Worker you have given them access to. Or, you can create an API token with the scoped access, which you can give to your agent to ensure they only have access to that one application. Hereâs an example of how to create an API token with permissions per Worker: Roles designed for how teams build When defining these roles, we wanted to strike the right balance. Overly broad roles force you to grant more access than intended, undermining the principle of least privilege, while providing too many individual permissions makes it difficult to know which ones to grant. We landed on four roles that reflect the levels of access you may want to give a person or agent: enough to debug a resource without exposing its content, read the content without changing it, make changes without being able to delete the resource, or fully manage it. We plan to use these same roles as we bring resource-level access controls to other Developer Platform products, including D1, R2, and KV. Each role can be applied at one of three scopes. For example, if you set the âmetadata read-onlyâ control, hereâs what that would look like at different levels: Developer Platform level: Access
Without proper controls, website owners have long faced a difficult tradeoff: allow your content to be used for AI training, or risk losing discoverability in search. That tradeoff exists because some of the largest organizations on the Internet use mixed-use crawlers: a single crawler serving both search and AI training. Refuse one, and you refuse the other. Today, Cloudflare is announcing a new Disallow AI Training setting that lets you easily stay indexed for search while refusing to let that same crawler train on your content. Apple, Google, and Microsoft honor or have committed (in a specified time frame) to honor this setting. Mixed-use crawlers were the hard part of the training question. AI Summaries are next. A site-wide yes or no is too blunt: how much of your content appears in a summary matters as much as whether it appears at all. An opt-out for AI summaries is already one of the requirements we've set for mixed-use crawler operators. By early next year, our goal is to let you control how much of your content is included â set once on Cloudflare, rather than with each operator separately. Why asking isnât enough Most site owners want to be found: by humans, agents, and (good) bots. But a significant portion of the open Internet is funded by advertising, subscriptions, or direct relationships with visitors, and those models only pay when someone actually arrives. Almost every site owner considers Search beneficial: less than 1% of Cloudflare sites choose to block Search bots. Training, however, is a different story: 17% of sites choose to enable some mechanism to block training. This is exactly why we decided site owners needed more granular controls, rather than a one-size-fits-all âBlock AI.â A robots.txt directive alone cannot solve this problem. Anyone can publish one, but it cannot identify who i
With artificial intelligence (AI) shrinking the window between vulnerability discovery and exploitation and lowering the barrier to entry for bad actors, new findings from Sysdig show that skilled human operators can move just as swiftly after gaining initial access. In one instance highlighted by the cloud security company, the threat actor pivoted from a vulnerable Marimo notebook to an SSH
Introduction Security teams have gotten pretty good at testing against what can hurt them. Can this EDR agent catch this payload? Will my organization fail the phishing simulation? Does this SIEM rule fire on this particular technique? And, in more mature organizations, this testing happens continuously rather than as a one-off exercise. But no matter how much you validate against these
Cybersecurity researchers have disclosed details of a mass-scanning campaign that has targeted Vite deployments siphon sensitive data. The first is an automated effort aimed at internet-exposed Vite development servers that's designed to steal cloud credentials, configurations from Amazon Web Services (AWS) and Microsoft Azure instances, and infrastructure state files, per F5 Labs. The
1Passwordâs FLAWED report , published on August 6, 2026, gives defenders a misleading picture of AI patching. Its headline says models produced clean fixes only 26% of the time. That figure includes experiments that deliberately instructed agents to apply the wrong fix, along with experiments in which agents could not compile or test their patches. The report risks making defenders less effective by discouraging them from using technology that could help them fix more vulnerabilities. Teams that take its headline at face value may leave repairable vulnerabilities unaddressed. We want our work to help defenders fix more vulnerabilities. This post shares real-world data on human and agent patch quality from our consulting projects and Patch the Planet. Weâre also releasing two agent skills: post-patch-validation to help agents test fixes, and review-walkthrough to help engineers review them. How the experiment produces a misleading headline Our review of 1Passwordâs code and data found four choices that make its 26% clean-fix rate a misleading guide to ordinary patching work. 1 The sample was selected for difficult fixes. The authors chose six vulnerabilities because their fixes were complex. Clean-fix rates ranged from 3% to 60% across those bugs, so
Cisco has warned that a new critical vulnerability impacting AsyncOS Software for Cisco Secure Email Gateway has come under active exploitation in the wild. The vulnerability, tracked as CVE-2026-76461, carries a CVSS score of 9.8 out of a maximum of 10.0. It has been described as a case of insufficient validation in the email parsing logic that could allow an unauthenticated, remote attacker
A Chinese threat actor has been attributed to a spear-phishing campaign that exploits recently patched security flaws in Google Chrome and Microsoft Windows to deliver a malicious JavaScript backdoor called GRIMWEDGE. Volexity, which is tracking the threat cluster under the moniker UTA0560, said the activity targeted multiple non-governmental organizations (NGOs) on September 1, 2026. "The
Monday, September 14
Florida Confirms DMV Breach as ShinyHunters Leak Exposes SSN Cards and Licenses Florida DMV confirms breach after ShinyHunters claimed access, while leaked files contain SSN cards, licenses, immigration records and government documents.
In the biggest-ever legal action against harmful deepfake websites, the Manhattan District Attorneyâs Office has seized 12 sites that collectively targeted around 1,200 victims.
An analysis of 160 deepfake websites reveals politicians in 22 countries appear on them. Nearly all of them are women.
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.
Sunday, September 13
Locating Flutter's TLS certificate verifier in a stripped libflutter.so, tested across 13 apps and 9 engine versions
Saturday, September 12
Plus: The US disrupts the internetâs biggest black market, a Conti ransomware hacker gets prison time, Meta fails to stop AI-generated videos of child abuse.
Friday, September 11
Presently sponsored by: Report URI: Guarding you from rogue JavaScript! Donât get pwned; get real-time alerts & prevent breaches #SecureYourSite I think what really resonates with me this week is being able to completely turn the tables on perceptions around things like AI being the big bad hacking tool the news would have you believe. There's the stat I talk about where it's had literally 0% impact on insurance payouts, the Dutch bloke who's now being tracked down based on his voice and Scott's MAG breach teardown that ultimately pointed to exposed keys rather than anything sophisticated. Then the mention in there about the Aussie headline where ChatGPT has hacked a court system down here when, in fact, a bloke used it to write a scraper he then pointed at the aforementioned website. It seems the truth just doesn't make for good headlines ââïž
VU#369611: ExLlamaV3 contains Denial of Service vulnerability via insufficient bounds checking on kernel dispatch index
Overview An out-of-bounds (OOB) memory access vulnerability involving unchecked array indexing has been identified in the exllamav3_ext compute unified device architecture ( CUDA ) extension. Successful exploitation can lead to an immediate denial of service or application instability. This vulnerability is tracked as CVE-2026-84286. Description An OOB memory access vulnerability exists in the exllamav3_ext module due to insufficient input validation. When the kernel parameter K is set to 0 in a crafted input, the extension generates a negative array index, resulting in a CUDA illegal memory access.s. The root cause is a missing bounds check in the kernel-table dispatch process. The checkpoint-derived block index ( cbi ) is used to access a fixed 24-entry array without confirming that either K or cbi fall within safe limits. Impact Primary impacts include Denial of Service (DoS) through process crashes and potential unstable execution states within applications utilizing the library. Solution The vendor has addressed this vulnerability in the main repository. Users are advised to update their installations or apply the fix from the merged pull request: https://github.com/turboderp-org/exllamav3/pull/310. Supply chain Downstream projects utilizing this library are indirectly exposed to this vulnerability. According to the ExLlamaV3 Dependency Graph, there are 49 total dependencies currently tracking this repository. Notable downstream projects directly affected by this
Synack's reporting experience now scopes to individual assets, generates AI-written executive summaries, and lets teams preview, save and share reports without leaving the platform. The post Executive-Ready Reporting Is Here: Whatâs New in the Synack Platform appeared first on Synack .
CISA has confirmed that ransomware operators are exploiting CVE-2025-14733, a critical remote-code-execution vulnerability affecting WatchGuard Firebox appliances. The bug can allow an unauthenticated attacker to execute code remotely under vulnerable configurations.
Today, weâre making Cloudflare CASB more powerful than ever by introducing automatic remediation policies. This means security teams can now design event-driven logic to revoke risky file shares and dispatch custom webhooks, without manual intervention. When we launched Cloudflare CASB , a cloud access security broker, we wanted to provide security teams complete visibility into the posture of their SaaS applications before misconfigurations became incidents. With a quick, clientless integration, CASB surfaces risks like overshared files, dormant admin keys and tokens, OAuth apps with excessive permissions â continuously, across users in the organization. For years, SaaS Security Posture Management (SSPM) tools such as Cloudflare CASB have functioned as a passive alarm system. Most SSPM tools tell you whatâs wrong but do not help you fix the issue, placing the burden on administrators to manage an ever-growing to-do list. A single misconfigured file-sharing policy across a Google Workspace tenant can generate thousands of findings in seconds, and even a disciplined team faces a window between detection and remediation measured in hours or days â more than enough time for a sensitive file to be downloaded, forwarded, or indexed. With automatic remediation policies, CASB customers can now configure the actions that should be invoked immediately after a new finding is identified. Shifting from reactive to proactive When we launched manual remediation actions earlier this year, we gave security teams the ability to resolve misconfigurations directly from the Cloudflare dashboard. This re
The new joint operation uses laser-based technology capable of detecting, tracking, and disabling commercial drones linked to human and drug trafficking.
I got hit by a fake Cloudflare popup and reverse engineered the malware it dropped. Here's what I found.
A few weeks ago I was browsing and got one of those "Verify you are human" Cloudflare popups. Except it wasn't Cloudflare â it was a ClickFix page that told me to press Win+R and paste a command. I'm on Linux so nothing executed on me, but I grabbed the file and spent time tearing it apart. Here's what's inside. **The file:** `Tourmaline.exe` (~10.8 MB) It's not a typical dropper. It's an Inno Setup 6.7.0 installer (Revision 2, 64-bit offsets â which broke every standard extraction tool I tried including innoextract) that bundles a full Python 3.11 runtime and two stages of obfuscated payload. **Stage 1 â Anti-sandbox time-lock** The first stage is a ~13 line obfuscated Python script with a clever trick: it counts down from 99,999,999 to find an XOR decryption key via brute force. On a sandbox with a 60-second timeout it never finishes. On a real machine it runs in seconds. I cracked it instantly using a known-plaintext attack on the first 4 bytes of the encrypted blob â recovered the key in O(1) without running the loop at all. **Stage 2 â Full Python RAT** Once decrypted, it's a hand-rolled Python backdoor with no external dependencies. Here's what it does: - **DNS tunneling** â all C2 traffic goes out as raw UDP DNS queries directly to `158.94.211.185:53`, disguised as `*.microsoft.com` lookups. Fully custom DNS packet builder, no library used. - **Blockchain dead-drop** â on startup it calls an Ethereum Sepolia smart contract (`0x2d7a04cca0c34005f58393f30ac725e25f19e5f5`) to get the current C2 IP, decrypted locally with a hardcoded ChaCha20 key. I verified this live â the contract returned the active IP. The attacker can update it any time, making IP blocklists useless. - **ECDSA-signed commands** â tasks from the C2 are P-256 signature verified before execution. You can't sinkhole it by intercepting the DNS tunnel. - **Arbitrary Python exec** â whatever the C2 sends back gets `exec()`'d in a persistent namespace. Full remote code execution. - **Persistence** as `TourmalineUpdate` in Task Scheduler, disguised as "Hardware monitoring service." Everything is in my GitHub repo â deobfuscated source, YARA rules, IOCs (JSON + CSV), and a decryption tool that reproduces the key recovery: **https://github.com/v-pun215/Tourmaline** Sample is included as a password-protected zip (`infected`) following the standard for sharing malware samples safely. Happy to answer questions on the methodology, especially the custom Inno Setup Revision 2 format parsing and the known-plaintext bypass.
Thursday, September 10
Back in 2021 I wrote a "how to hack APIs" blog for Detectify that got loads of traction, but it's way out of date now. I rewrote it for 2026 to include a bunch of the newer popular technologies and AI.
A novel technique for writing ELF shared objects and achieving code execution in Python, Ruby, and Node.js using the sqlite\_dbpage virtual table.
1.1.1.1 now validates DNSSEC signatures made with ML-DSA-44 , a post-quantum signature algorithm standardized by the National Institute of Standards and Technology (NIST) . This is a first step toward preparing DNSSEC for a future in which todayâs signature algorithms are no longer secure. Cloudflare plans to achieve full post-quantum security by 2029 . Much of the work so far has focused on TLS, but public-key cryptography is used in many other systems, including DNSSEC. While we began experimenting with post-quantum key agreement in TLS in 2019 and enabled support for all customers in 2022, post-quantum signatures have not yet received comparable testing in DNSSEC. There is also some urgency. Widespread client adoption of post-quantum TLS took years, partly because larger messages exposed assumptions and bugs in existing network software. That experience showed why early large-scale testing matters. We cannot wait until quantum computers become an immediate threat. The problem is that post-quantum signatures are large. Each ML-DSA-44 signature is 2,420 bytes, exceeding common DNS-over- UDP limits before the response includes anything else. At the same t
US government adviser Paul Christiano warns of risks to AI industry as he joins OpenAIâs non-profit foundation OpenAI is not on track to reduce the risk of âcatastrophicâ loss of control to an acceptable level, a member of its non-profit board has said, amid spreading public and political concern that super-advanced AIs could one day wipe out humanity. Paul Christiano, a US government technology adviser, said : âThere is a meaningful risk that rapid acceleration in AI capabilities leads to catastrophic and irreversible loss of control in the very near term.â Continue reading...
InquiryIQ, a previously unreported prototype, tested a model from xAI, maker of Grok, to surface associates, social accounts, and other information about people identified through Clearview.
Wednesday, September 9
The City Attorneyâs Office has asked Meta to explain how the harmful ads repeatedly ran on Facebook and Instagram. The company claims the ads are not under the cityâs jurisdiction.
The new Apple Watch includes several âintelligentâ listening features that have privacy and security baked in. But the protections canât change the facts of what the tools do.