Door categorieën bladeren
Ontdekken
Fiverr Pro
Nederlands
$
USD
I will do llm penetration testing, test ai chatbot security
Offensive Security Specialist, AI security Tester, Virtual Assistant
I’m Shakil Ahmed Srabon (Shahaz), I work in offensive security, testing AI systems for weaknesses, and I also handle virtual assistant work across tech and business operations. I've spent 5 years in t...
Over deze dienst
An open chat window means anyone can try to manipulate your application logic.
I operate as your AI Security Specialist and Pentester. Combining 5 years in tech with 2 years in offensive security, I simulate real-world attacks to ensure your custom LLM integrations, AI agents, and web applications cannot be manipulated or exploited.
What I Audit & Secure:
- Prompt Injection: Testing boundaries to block unauthorized system bypasses.
- Jailbreaking Flaws: Hardening instructions so your bot never breaks character.
- Data Leak Isolation: Ensuring the AI never reveals backend API keys or user logs.
- Insecure Output: Preventing the model from executing malicious code on your server.
- API & Web Vulnerabilities: Checking backend infrastructure for OWASP, XSS, and SSRF flaws.
What You Get:
- OWASP Alignment: Every audit maps directly to the official OWASP LLM Top 10 framework.
- Zero Overhead: No extra tool costs. I use my own infrastructure to run simulations safely.
Besides, you receive a comprehensive, structured PDF audit report containing clear proof-of-concept findings alongside precise, developer-ready mitigation patches.
Check my FAQs and message me right now to secure your AI system!
Mijn portfolio
Veelgestelde vragen
What is included in the Basic Package?
I do not offer basic or superficial testing. Even at this tier, you receive an advanced-level security audit targeting frontend prompt injection and jailbreak vulnerabilities on your chatbot. You get a fully detailed, professional PDF report containing clear patch instructions.
What is included in the Standard Package?
This expands to a comprehensive deep-logic security assessment mapping to the OWASP LLM Top 10 framework. I test for prompt chaining, data leaks, and web vulnerabilities (XSS/SSRF), delivering a heavily detailed PDF audit report with proof-of-concept exploit findings.
What is included in the Premium Package?
This is a complete enterprise-grade penetration test of your entire AI ecosystem. I test frontend interfaces, backend APIs, data layers, and agentic workflows for severe code flaws. You receive an exhaustive PDF audit report alongside direct developer mitigation support.
How is this different from standard website penetration testing?
Standard firewalls and web pentests look for traditional software bugs. They completely miss semantic exploits. Because I specialize in LLM behaviors, I test how your AI processes human text, preventing hackers from using trick prompts to control your application logic.
What happens if a complex AI bug takes longer to fix or test?
For me, system security is the priority, and money is secondary. If a critical vulnerability requires deeper, complex testing beyond our package bounds, I will gladly invest the extra time for free. I am here to deliver true operational safety, not watch a clock.
Will my API keys, training data, and system endpoints be safe?
Data protection is paramount. With 2 years in offensive security, I practice strict asset isolation. I use a secure local testing environment, sign NDAs upon request, and never store or expose your sensitive training datasets, proprietary prompts, or API credentials.
Do you run these security attack simulations on my live production website?
To ensure zero business downtime, I highly recommend testing on a staging or development environment. However, if testing a live chatbot is mandatory, I use controlled, non-destructive attack vectors that safely isolate vulnerabilities without disrupting your active users.
Are there any hidden fees or extra tools I need to purchase?
There are no hidden costs. I use my own advanced exploit tools and secure attack infrastructure to run all simulations. You only need to provide the target access permissions or endpoint URLs necessary for me to execute the penetration test safely.
Do we use Zoom or screen sharing to discuss the technical layout?
Yes! Complex AI system architectures require clear alignment. We can easily hop on a live video or screen-share call right through Fiverr’s secure platform to review your infrastructure maps, look at system logs, and ensure we map out the correct security boundaries
