TryHackMe (@tryhackme) header
TryHackMe (@tryhackme) profile picture

An online platform that makes it easy to break into and upskill in cyber security, all through your browser.

Joined November 2018
Photo by TryHackMe (@tryhackme)Photo by TryHackMe (@tryhackme)Photo by TryHackMe (@tryhackme)Photo by TryHackMe (@tryhackme)Photo by TryHackMe (@tryhackme)Photo by TryHackMe (@tryhackme)Photo by TryHackMe (@tryhackme)Photo by TryHackMe (@tryhackme)Photo by TryHackMe (@tryhackme)Photo by TryHackMe (@tryhackme)
TryHackMe retweeted
We tested 11 AI models on 120 vulnerabilities to see which one finds the most. Claude Mythos 5.2 came first at 86%. Claude Opus 5.5 found 80%, with half the false alarms and half the cost per finding.
3
3
1
13
2,109
TryHackMe retweeted
New Windows Incident Response & Forensics Learning Path! This is a BEAST of a release: 130 hours, 13 modules & 69 rooms. Built around real threat actors and it covers more than 100 MITRE ATT&CK techniques seen in actual incidents! The path follows the full attacker kill chain in the order a real incident unfolds: from initial access and execution, through persistence, credential access, privilege escalation, domain compromise, lateral movement, and on to exfiltration and impact. You’ll look at each incident from both an offensive and defensive perspective - act as the hacker, then respond. This is intended for those with existing IR skills, but you can work towards it by completing our SOC Level 2 and Advanced Endpoint Investigations paths first. Live now on @tryhackme 🫡 tryhackme.com/path/outline/w…
1
4
18
2,169
TryHackMe retweeted
Almost 50% of AI-generated code is vulnerable... yup you read that right... It means most companies are introducing MORE vulnerabilities than EVER, at a time when it’s never been easier for threat actors to pentest at scale. Especially scary given they’re shipping more than ever, and few actually read what AI is generating. To explain the attached image: the line graph on the right shows that from 2023 to 2026, syntax errors (grey line) improved a lot, while secure code generation only marginally improved (red line). The table on the left shows the security pass rate (higher is better). Veracode, who put the report together, tested code snippets generated by different models to see how much of it was vulnerable. Some vulnerability categories are far more common than others. e.g. SQLi had a high security pass rate at 83% (bc its easier to identify just by code), but XSS' pass rate was 15%. It means we'll see more of one vulnerability category over time. In fact, even prompting "make this more secure" or "fix X vulnerability" in many cases found that it made the app less secure. This means that: 1️⃣ More vulns are being shipped than ever 2️⃣ AI can’t reliably fix its own security flaws 3️⃣ You need to test at both the code and app level 4️⃣ And test regularly, triggered by critical actions like deploys. NoScope does exactly this. I'll link the research in the comments - including a presentation that goes over this in more depth we gave a few weeks ago. When I heard almost 50% of AI-generated code was vulnerable, I didn’t believe it. So I read the research and tested it myself.. @tryhackme
5
6
19
2,816
TryHackMe retweeted
𝗗𝗼𝗻'𝘁 𝗴𝗲𝘁 𝗹𝗲𝗳𝘁 𝗯𝗲𝗵𝗶𝗻𝗱.. every company has AI agents as a new attack surface but people are TOTALLY UNAWARE how to exploit them beyond basic "prompt injection" - like agent hijacking, tool & identity abuse, exploiting code execution, memory poisoning, and more. Our new learning path "Agentic AI Security" covers all of this with real agents. This is NOT a simulation. Our engineering team built new tooling so you can deploy real agents, on current models, in the infrastructure companies actually use today! The path covers the full (new) OWASP Agentic Top 10, from ASI1 to ASI10 - and includes both offensive AND defensive techniques. Live now on @tryhackme 🫡
8
3
17
1,736
TryHackMe retweeted
We’re hosting a live webinar next Thursday, 22nd Oct at 9:00 AM EDT / 2:00 PM BST, on Hacking & Securing Apps Where Devs Ship AI Written Code. @springjben and @KevalJagani1 will cover the vulnerabilities AI tends to introduce, why it happens, and how to secure them. Register now, link in comment 👇
1
3
1
6
1,712
From solving TryHackMe rooms to meeting the community in person. Attended the first-ever TryHackMe Community Gathering in Delhi! Great conversations, new connections, and knowledge shared with fellow cybersecurity enthusiasts. Thanks to everyone who made it happen! @tryhackme
1
1
3
1,434
TryHackMe retweeted
Jira integration is now live on NoScope. Every finding now syncs directly to the Jira dashboard, with severity, evidence and remediation steps attached to each ticket.
2
5
1,310
With @trynoscope, you’ll soon be able to watch our AI pentester work in real time. See how the agent explores your web app, builds an attack graph, investigates potential vulnerabilities, and validates what it finds. You’ll have visibility into: • The agent’s reasoning and investigation • Every request it sends • The attack path it takes • Vulnerabilities it discovers and how it validates them • Attack classes including Recon, XSS, SSRF, IDOR, access control, and more AI pentesting shouldn’t be a black box. You should be able to see exactly what the agent did and how it found the vulnerability.
4
6
1
17
3,125
TryHackMe retweeted
Replying to @tryhackme
@tryhackme awesome seeing you at defcon 34
1
19
2,734
TryHackMe retweeted
We asked a Gitea server for a token we had no business getting. It said 200 OK. Here's how NoScope's AI agent found a 4-year-old flaw that exposed private container images on ~30,000 servers 🧵
3
6
1
20
2,655
TryHackMe retweeted
Prompt engineering is a really interesting topic ngl #AISecurity @tryhackme
2
2
12
2,391
TryHackMe retweeted
We're adding a badge for solving the NoScopeRCE room on TryHackMe. 2500+ people have solved it already, if you haven't, now's the time. Craft a sandbox escape payload using the returnClass binding, trigger it through the Extensions API, and chain it into a full reverse shell. No setup needed. Also see how NoScope autonomously found and validated CVE-2026-35482, and how to run it against a live target. Solve it, grab the badge, link in comments
1
4
17
3,367
TryHackMe retweeted
We recently found a critical RCE, and turned it into a free @tryhackme room. 1500+ people have already solved it. If you haven't tried it yet, it's worth a go. You'll craft a sandbox-escape payload using the returnClass binding, trigger it through the Extensions API, and upgrade to a full reverse shell, hands-on with no setup required. You'll also see how NoScope autonomously identified and validated CVE-2026-35482, and how to configure it to run a full automated pentest against a live target. Give it a try and let us know what you think. Link in the comments
1
3
10
2,537
TryHackMe retweeted
Last week we hosted webinar on, "AI is Accelerating Attacks: How to Evolve Your Pentest Methodology" with 1,400 people signed up. Thank you to everyone who joined! If you missed it (or want to rewatch), the full recording is up now: noscope.com/webinars/pentest… We covered how attackers are using AI to move faster than traditional security teams can respond, and why you should be evolving your pentest methodology in 2026.
1
4
12
1,856
A quick update on the Gitea vulnerability. Over 30,000 instances may still be exposed. The flaw (CVE-2026-27771) lets anyone pull private container images without authentication. Many organizations have not patched yet. If you run Gitea, check your version immediately and update to 1.26.2 or later. Temporary workaround: Enable REQUIRE_SIGNIN_VIEW.
🚨 Gitea flaw exposes private container images without authentication. thehackernews.com/2026/05/gi… CVE-2026-27771 affects all Gitea versions before 1.26.2 and likely impacts 30,000+ deployments worldwide. Attackers can pull private images without an account or password. Update now or enable REQUIRE_SIGNIN_VIEW as a temporary workaround.
7
22
1
72
19,164
TryHackMe retweeted
For 4 years, any person on the internet could pull private container images from 30,000+ Gitea deployments. No account. No password. No credentials of any kind. NoScope discovered it. CVE-2026-27771. Here's the full story. 🧵
4
5
2
21
5,806
TryHackMe retweeted
Replying to @securestep9
@securestep9 will be joining the panel discussion at @tryhackme CTF event tomorrow in London! 🎉 Be sure to ask Sam more about OWASP and how you can get involved. #AI #CTF #community #opensource #appsec
Meet the speakers 🤝 Builders, creators & cyber leaders all joining us live in London 🇬🇧 Can’t make it in person? We’re streaming everything until 12PM: 🎤 Talks, 🛰️ panel & the ⚔️ live challenge. Livestream link will drop on the landing page👇 ▶️ tryhackme.com/ai-odyssey?utm…
1
3
14
3,862
TryHackMe retweeted
For anyone using Vercel, a few important steps to take after today’s incident: 1. Rotate your secrets 2. Move away from long-lived credentials 3. Review audit logs for any unusual access We’re offering a free pentest for Vercel users. Reach out to get started.
We’ve identified a security incident that involved unauthorized access to certain internal Vercel systems, impacting a limited subset of customers. Please see our security bulletin: vercel.com/kb/bulletin/verce…
3
5
27
9,185