Course Overview
Generative AI is transforming the software industry, with tools like GitHub Copilot and Codeium enabling developers to achieve unprecedented levels of efficiency. While this is exciting progress, it also raises important concerns, encouraging stakeholders to approach these technologies with care. Current AI tools often lack the nuanced understanding necessary to address subtle, yet critical aspects of software development, particularly in the domain of security.
This course provides a comprehensive insight into the responsible use of generative AI in coding. Participants delve into topics in software development that are most likely to be impacted by careless use of generative AI, including authentication, authorization, and cryptography. The curriculum also includes an analysis of how AI tools like Copilot handle secure coding practices related to key vulnerabilities outlined in the OWASP Top Ten, such as path traversal, SQL injection, or cross-site scripting.
Through hands-on learning and experimenting, participants will get a solid understanding of both the strengths and limitations of AI-assisted development. In addition, case studies of real-world incidents showcase the consequences of insecure code and demonstrate the dual nature of generative AI as both a resource and a potential risk.
By the end of the course, developers will be equipped with the knowledge and skills to integrate AI tools into the software development lifecycle responsibly, enhancing efficiency without compromising security or product quality.
Who should attend
C/C++ developers using Copilot or other GenAI tools
Prerequisites
General C++ and C development
Course Objectives
- Understanding the essentials of responsible AI
- Getting familiar with essential cyber security concepts
- Correctly implementing various security features
- Identify vulnerabilities and their consequences
- Learn the security best practices in C++
- Managing vulnerabilities in third party components
- Input validation approaches and principles
- All this put into the context of GitHub Copilot
Wrap up
- Secure coding principles
- Principles of robust programming by Matt Bishop
- Secure design principles of Saltzer and Schroeder
- And now what?
- Software security sources and further reading
- C and C++ resources
- Responsible AI principles in software development
- Generative AI – Resources and additional guidance
Course Content
Day 1
Coding responsibly with GenAI
- What is responsible AI?
- What is security?
- Threat and risk
- Cyber security threat types – the CIA triad
- Cyber security threat types – the STRIDE model
- Consequences of insecure software
- Security and responsible AI in software development
- GenAI tools in coding: Copilot, Codeium and others
Memory management vulnerabilities
- Assembly basics and calling conventions
- x64 assembly essentials
- Registers and addressing
- Most common instructions
- Calling conventions on x64
- Calling convention – what it is all about
- Calling convention on x64
- The stack frame
- Stacked function calls
- Buffer overflow
- Memory management and security
- Buffer security issues
- Buffer overflow on the stack
- Buffer overflow on the stack – stack smashing
- Exploitation – Hijacking the control flow
- Lab – Buffer overflow 101, code reuse
- Exploitation – Arbitrary code execution
- Injecting shellcode
- Lab – Code injection, exploitation with shellcode
- Case study – Stack BOF in FriendlyName handling of the Wemo Smart Plug
- Pointer manipulation
- Modification of jump tables
- Overwriting function pointers
- Best practices and some typical mistakes
- Unsafe functions
- Dealing with unsafe functions
- Lab – Fixing buffer overflow (exploring with Copilot)
- Using std::string in C++
- Manipulating C-style strings in C++
- Malicious string termination
- Lab – String termination confusion (exploring with Copilot)
- String length calculation mistakes
Day 2
Memory management hardening
- Securing the toolchain
- Securing the toolchain in C++
- Using FORTIFY_SOURCE
- Lab – Effects of FORTIFY
- AddressSanitizer (ASan)
- Using AddressSanitizer (ASan)
- Lab – Using AddressSanitizer
- Stack smashing protection
- Detecting BoF with a stack canary
- Argument cloning
- Stack smashing protection on various platforms
- SSP changes to the prologue and epilogue
- Lab – Effects of stack smashing protection
- Runtime protections
- Runtime instrumentation
- Address Space Layout Randomization (ASLR)
- ASLR on various platforms
- Lab – Effects of ASLR
- Circumventing ASLR – NOP sleds
- Circumventing ASLR – memory leakage
- Non-executable memory areas
- The NX bit
- Write XOR Execute (W^X)
- NX on various platforms
- Lab – Effects of NX
- NX circumvention – Code reuse attacks
- Return-to-libc / arc injection
- Return Oriented Programming (ROP)
- Protection against ROP
- Case study – Systematic exploitation of a MediaTek buffer overflow
Day 3
Common software security weaknesses
- Security features
- Authentication
- Password management
- Inbound password management
- Storing account passwords
- Password in transit
- Lab – Is just hashing passwords enough?
- Dictionary attacks and brute forcing
- Salting
- Adaptive hash functions for password storage
- Password policy
- NIST authenticator requirements for memorized secrets
- Password database migration
- Code quality
- Code quality and security
- Data handling
- Type mismatch
- Lab – Type mismatch (exploring with Copilot)
- Initialization and cleanup
- Constructors and destructors
- Initialization of static objects
- Lab – Initialization cycles (exploring with Copilot)
- Unreleased resource
- Array disposal in C++
- Lab – Mixing delete and delete[] (exploring with Copilot)
- Object oriented programming pitfalls
- Accessibility modifiers
- Are accessibility modifiers a security feature?
- Inheritance and object slicing
- Implementing the copy operator
- The copy operator and mutability
- Mutability
- Mutable predicate function objects
- Lab – Mutable predicate function object
Using vulnerable components
- Security of AI generated code
- Practical attacks against code generation tools
- Dependency hallucination via generative AI
- Case study – A history of GitHub Copilot weaknesses (up to mid 2024)
[/list]