Ethical Hacking Gen AI/LLM Complete Hands on

Explore artificial intelligence and its subfields—AI, ML, deep learning, and generative AI—through real-world examples like Siri, Alexa, spam filters, autonomous driving, and ChatGPT.

Explore what AI is and how it benefits automation, personalization, efficiency, decision making, and innovation through data-driven insights and automated processes.

Explain what generative AI is and its benefits, including personalized recommendations and improved healthcare, with examples from Netflix, IBM Watson Health, ChatGPT, and Dalle A.

Learn what prompts are and how they drive responses in large language models, then create clear prompts with topic, context, and specific details, using practical examples for support and translation.

Explore how prompt injection compromises large language models by crafted inputs that trigger unintended actions. See examples like prompts that override instructions and draft favorable contracts to understand security implications.

Learn how prompt injection exploits vulnerable inputs to manipulate language models, enabling malicious prompts to bypass instructions, exfiltrate data, escalate privileges, and trigger destructive actions.

Explore prompt injection types, including direct prompt injection (jailbreaking) and indirect prompt injection, and learn how attackers manipulate inputs and systems interacting with the large language model.

Learn practical steps to prevent prompt injection in smart AI applications by using input sanitization, fixed prompt templates, and training data to recognize and ignore malicious prompts.

Explore prompt injection in AI chat bots using infographics, showing how attackers craft malicious prompts to trick LLMs, retrieve data, and how to prevent such attacks to keep data safe.

Explore prompt injection vulnerabilities in large language models through an interactive lab based on OWASP top ten, illustrating how crafted inputs can expose the admin password and threaten data security.

Explore the OWASP top ten vulnerabilities in large language models through an interactive prompt engineering lab, featuring hands-on exercises and level-based challenges to test prompt effectiveness and security awareness.

Examine a case study of prompt injection vulnerabilities in a healthcare AI assistant powered by an LLM, and perform a hands-on lab to exploit and secure against such attacks.

Explore insecure output handling in language models, showing how unvalidated llm output like JavaScript or markdown can trigger csrf, ssrf, privilege escalation, or remote code execution.

Identify vulnerable input fields and inject malicious prompts to trigger insecure output handling in AI systems. See real-world scenarios where cross-site scripting via prompts leads to account takeover.

Explore insecure output handling in large language models, reveal attacker steps from input manipulation to executing malicious prompts, and learn safeguarding strategies to prevent real-world threats.

Apply a zero trust approach to model interactions by validating responses to backend functions and encoding outputs to users, following the OWASP ASVS guidelines for input validation and output encoding.

Identify vulnerable input fields in an llm-powered application, inject html and javascript, and observe how insecure output handling enables code execution and data exposure in the lab.

Explore how training data poisoning corrupts AI models by inserting biased or malicious data, causing incorrect predictions and misidentifications while highlighting data integrity risks.

Learn how training data poisoning happens when attackers insert malicious documents and prompts to corrupt a model during pre-training, producing inaccurate responses in healthcare and fraud detection.

Understand how training data poisoning can mislead large language models by injecting manipulated inputs, producing poisoned responses and undermining user trust and system reliability.

Explore defenses against training data poisoning in large language models, including supply chain verification, ML BOM attestations, data source legitimacy, and secure pipelines with sandboxing and monitoring.

Explore how training data poisoning injects malicious data into an LLM's training set to influence responses. Show how attackers exploit vulnerabilities and reveal data poisoning risks, underscoring secure data pipelines.

Explore model denial of service attacks by examining how overwhelming AI systems with too many requests or complex tasks can exhaust computational resources, slow down, or crash services.

Explore how model denial of service occurs when attackers flood llms with requests, overwhelming the backend, slowing responses, and triggering outages in online retail scenarios.

Explore how model denial of service occurs with infographics, as attackers flood an LLM backend with resource-intensive queries, causing crashes and service interruption.

Prevent model denial of service in llms and chatbots. Implement rate limits, input validation and sanitization, resource allocation, and query complexity controls, with monitoring and alerts for malicious requests.

Explore how model denial of service disrupts LLM powered applications by flooding requests, causing 500 errors and degraded performance, and learn defenses to protect availability and reliability.

Examine supply chain vulnerabilities that affect large language models and chatbots, including data sources, algorithms, hardware, and human expertise, and how compromised updates threaten security and reliability of these systems.

Examine how a single supply chain failure affects a banking application using an LLM, from vulnerability entry and malicious data to unauthorized transfers.

Explore supply chain vulnerabilities of lm systems, including backdoors in third party libraries and plugins, compromised data sets and pre-trained models, and how attackers exploit these risks, with infographics.

Explore real-world supply chain attacks, and implement practical steps to secure AI systems and LLMs, including regular audits, data verification, secure coding, and vendor management.

Explore supply chain vulnerabilities in AI components and dependencies, demonstrating how tampering with third party libraries can trigger unauthorized actions and data breaches in real world apps.

Learn how large language models can disclose private information and how to safeguard data when using AI chatbots, including risks from unsanitized responses and prompts that enable data exfiltration.

Explore how AI models disclose sensitive information and the mechanisms behind it, using a banking chatbot step-by-step example, and learn privacy safeguards to prevent leaks.

Identify how attackers probe banking chatbots for sensitive data and visualize a step-by-step infographic to protect user information and improve ai security.

Implement data anonymization to remove sensitive data during training and anonymize identifiers; enforce role-based access controls, conduct regular audits, and monitor logs in real time to prevent data leaks.

Explore how prompt injection leads to sensitive information disclosure in AI systems, shown by a banking chatbot case study that leaks credit card details and privacy breaches.

Identify insecure plugin design vulnerabilities in large language model plugins to secure ai applications. Emphasize input validation and avoiding plain text storage to prevent injection and ssrf attacks.

Explore how insecure plugin design arises from weak security measures and lack of input validation, enabling attackers to exploit plugins and access credit card details in LM based apps.

Explore insecure plugin design in ai-driven applications and how attackers exploit chatbots through weak security and malicious input, triggering unintended commands or accessing restricted data.

Implement secure design practices to reduce plugin vulnerabilities and boost your application's security. Regular security audits, strict input validation, secure API design, and TLS encryption safeguard plugins.

Explore how insecure plugin design enables attackers to exploit llm plugins processing untrusted inputs and weak access controls, and learn how to secure plugins in this hands-on lab.

Explore excessive agency in large language models and its security risks, including unauthorized transactions, and learn safeguards to align llm behavior with user intent.

Explore how excessive agency arises in large language models by examining attacker prompts, model responses, and the need for user input control to prevent autonomous actions.

Explore excessive agency in language models with simple infographics, showing how attacker-crafted prompts can drive unintended actions, and how to prevent such abuse in practice.

Explore strategies to prevent excessive agency in large language models by applying strict authentication, contextual awareness, access control, and audit trails to ensure safe, accountable ai behavior.

Examine how excessive agency in llm-powered apps enables attackers to manipulate actions, access user data, and degrade trust, reliability, and privacy, underscoring methods to limit ai's decision making.