
There are 10 types of people in the world. Those who understand binary, and those who don't. All opinions and views are my own. #BsidesDub organizer
Naples,Italy
Joined September 2007
- Tweets273K
- Following13.6K
- Followers12.4K
- Likes78.5K
Securityblog retweeted
👋 Did you know you can nominate anyone to be a guest on Entra Chat? Even yourself!
Just sign in and submit at my.admin.news/me/guests
Bonus: If you've done a migration from MIM to Entra ID Governance, I would love to know.
Securityblog retweeted
My Apple exploit chain leveraged 7 different bugs and multiple techniques to achieve reliable code execution. I'm glad to see none of them were duplicates.
Securityblog retweeted
Fireworks is SOC 2 and ISO 27001/27701/42001 compliant. They’re still getting hacked. These certs are just security theater at this point.
Companies pay tens/hundreds of thousands a year for these audits because their customers assume it means their data is safe. And clearly, it does not.
Before we had cyber capable AI, it was much less obvious. But now it just means you had like $12,000 to pay a compliance shop and you have the bare minimum in place.
It’s time for an actually comprehensive framework that, you know, works.
Fireworks, the most highly-valued out of a growing group of inference providers, told customers including Harvey that it had been hit with a hack that exposed customer data. Some customers are now considering moving their business off the platform.
theinformation.com/articles/…
Securityblog retweeted
There is no excuse anymore. You can’t hide behind SOC2.
Running continuous red teaming against your company is now:
A. Economical
B. Absolutely critical
Red teaming is just truth seeking about the state of your security from first principles.
SOC 2 is reasoning by analogy that you should be secure.
Fireworks is SOC 2 and ISO 27001/27701/42001 compliant. They’re still getting hacked. These certs are just security theater at this point.
Companies pay tens/hundreds of thousands a year for these audits because their customers assume it means their data is safe. And clearly, it does not.
Before we had cyber capable AI, it was much less obvious. But now it just means you had like $12,000 to pay a compliance shop and you have the bare minimum in place.
It’s time for an actually comprehensive framework that, you know, works.
Securityblog retweeted
so ..now the bugbounty hunters have become triagers ..triaging the findings of their AI agents and submitting to the bugbounty platform where, now, a AI-Triage agent will check your report before a human triager sees your report...
AI --> Bugbounty hunter --> AI --> Triager
Securityblog retweeted
US security firm N-able pulls the plug on N-central, its key platform: cnews.link/n-central-offline…
Securityblog retweeted
Interested in old video game vulnerabilities? Check out our latest blog post on 𝘚𝘵𝘢𝘳 𝘞𝘢𝘳𝘴 𝘌𝘱𝘪𝘴𝘰𝘥𝘦 𝘐: 𝘙𝘢𝘤𝘦𝘳 by @tomtombinary 👇
synacktiv.com/en/publication…
On 2 November 2026, two Administrator sessions on an internet-exposed Copeland XWEB Pro 1.9.0 refrigeration controller read settings across 31 devices. We examine the event log and its limits, and set out how operators can reduce exposure.
ransom-isac.org/blog/the-cop…
Securityblog retweeted
‼️ BREAKING: N-able has told customers to shut down their software IT firms use to remotely run their clients' computers, after 33 alleged zero-days in it went public.
It has also pulled its own cloud version of the software, N-central, offline. One hacked server could give attackers a path into every computer it manages.
An account calling itself Oracle-Security posted the flaws under the name "Able2Shell", security firm Field Effect reports.
N-able says it has not seen them exploited. There is no patch yet.
Securityblog retweeted
France formally attributes cyberattacks on refrigeration controllers to Iran's Revolutionary Guards. The goal was destruction - sabotage. Food worth tens of millions of euros spoiled and at least 15 fires broke out. Attackers also went after Schneider Electric equipment at hydroelectric plants, though with limited effect.
Securityblog retweeted
I haven't seen anyone post them actually doing attacks and defence with ARTEX etc.
I've seen a lot of people TALK.
I've seen people claim things that don't match up to my experience in my lab/testing/use of LLMs or ARTEX specifically.
1
13
1,664
Securityblog retweeted
Beyond ARTEX: 45 Agentic Attack Frameworks, What They Hunt For, and How to Find Them
blog.gitguardian.com/ai-pent…
Securityblog retweeted
Threat Hunting Labs’ next League Season 2026 competition runs 11–13 December (UTC): From Azure Identities to Patient Records.
You join the response team for an organization running on Microsoft Azure. There’s an alert and the tenant’s logs. Your job is to work out who got in, how they did it, and what they took, then rebuild the intrusion from start to finish.
You’ll investigate Entra ID sign-ins and audit, Azure Activity, Microsoft Graph, Blob Storage, Azure SQL audit, Windows VM event logs and virtual network flow logs. Query them in the browser using THL’s Search Console with KQL.
There are 57 questions across threat hunting and incident response, worth 562 points. Pick your own start time during the competition window. You get one individual timed run of up to four hours, with a ranked leaderboard.
Entry is USD 15 or eight credits for THL subscribers. Registration closes 13 December at 19:49 UTC.
Register once the page goes live:
threathuntinglabs.com/compet…
Securityblog retweeted
SEGA has confirmed a data breach exposing 2.2 million newsletter subscribers’ email addresses after its third-party provider, ARARA, was hacked.
SEGA says no names, passwords, payment details, phone numbers or home addresses were exposed. Regular SEGA Account email addresses were not affected.
ARARA’s wider incident affected 77 million email addresses across nearly 380 companies.
The company also clarified that this incident is unrelated to the separate cyberattack that took SEGA Fave's ALLNet and Aime arcade services offline.
Securityblog retweeted
ARTEX, an AI-native autonomous penetration framework tied to Chinese 🇨🇳 Storm-0940, was observed attacking live AWS environments in October 2026, hunting for financial data, secrets, and cloud credentials through compromised IAM keys and public Cognito pools.
Key findings:
- ARTEX is a hosted, LLM-orchestrated attack platform published on GitHub in July 2026, reaching 12,000 forks before going closed-source on October 8. Hundreds of internet-exposed consoles were found, most presenting a Chinese-language login page titled "ARTEX - Autonomous Penetration Testing Console," with infrastructure concentrated in Singapore 🇸🇬 and Hong Kong. Microsoft attributed it to Storm-0940. The platform supports OpenAI, Anthropic, DeepSeek, and compatible gateways, rotating across API keys and providers if one is blocked, so a single campaign can span multiple inference providers.
- Two initial access paths were observed in the wild. First: compromised long-term AKIA IAM access keys. Once in, ARTEX's dominant action was FilterLogEvents with a consistent limit of 1000, searching CloudWatch for financial institution names combined with terms like Fund, Invoice, Billing, and LLC, then widening to queries for "s3://", "presigned", ".pdf", ".xlsx", ".csv", AKIA strings, internal IP ranges (172., 10.), passwords, and MongoDB references. Activity on October 20 originated from 103.54.154[.]40, an IP resolving to zeofast[.]cn.
- Second path: public Amazon Cognito user pools with anonymous sign-up enabled. A spike on October 26-27 from 154.64.252[.]77 showed ARTEX iterating through PreSignUp Lambda validation errors (including attempts using api.github[.]com and example[.]com as tenant domains), then completing the full flow: SignUp, InitiateAuth, RespondToAuthChallenge, GetUser, GetTokensFromRefreshToken. It then attempted broad read-only enumeration: ListBuckets, ListSecrets, DescribeParameters, ListFunctions20150331, GetRestApis, and more under the CognitoIdentityCredentials role session name. In this case, the authenticated role lacked permissions and all calls failed.
- Every session reviewed opened with unauthenticated S3 requests appearing in CloudTrail with userIdentity.type of AWSAccount and accountId of anonymous. ARTEX also performed pre-fuzzing baseline requests using placeholder keys like zzz_nonexistent_abc.txt and Notexist-xyz before bursts of S3 enumeration. A request for a clearly nonexistent object name immediately before a guessing burst is a detectable behavioral signal. User agents are inconsistent but include artex-readonly/1.0, artex-pentest/1.0, artex-spawn/1, ARTEX-worker/1.0, and Mozilla/5.0 ARTeX, among others listed in the report's IOC table.
Practitioner takeaway: In CloudTrail, hunt for FilterLogEvents calls with limit=1000 at volume, anonymous S3 access preceding credential use, and CognitoIdentityCredentials session names followed by broad Get*/List* enumeration. Cross-reference against the IP list in the report, which includes 103.54.154[.]40, 154.64.252[.]77, 154.36.136[.]197, 154.36.136[.]23, 82.158.88[.]233, and others. Flag any S3 request for objects matching zzz or notexist patterns immediately before an enumeration burst. The report's IOC table covers 10 IPs and 10 user agent strings. Rotate or remove long-lived AKIA keys: they are ARTEX's preferred foothold.
#DFIR_Radar
Securityblog retweeted
If it wasn’t already obvious, please dont try to run ARTEX yourself.
We have already talked to at least 10 companies last week alone that had caused real damage to their orgs because they thought putting abliterated deepseek in artex was a good idea.
Securityblog retweeted
Automating AI Hacking with Tree of Attack - John Stigerwalt @WKL_cyber
whiteknightlabs.com/2026/09/…
Securityblog retweeted
My first internal pentest ever was as wild as it gets…
This is a snippet of my chat with Michael Lisi on his podcast Hack Dissection.
Watch it here: youtu.be/p_GaketDENI?is=MgOY…
Thanks to Mike for having me on!
3
16
1,579
Securityblog retweeted
Forest Blizzard’s Frost Armada network hijacked DNS settings on compromised routers to harvest credentials. Danny Adamitis explains how nation-state actors react to public disclosure at #CYBERWARCON 2026.
Nov 20 | Arlington, VA + virtual
cyberwarcon.com








































