Backdoor Attack
An attack that embeds a hidden trigger in a machine learning model during training, causing it to produce attacker-specified outputs when the trigger is present in the input.
In plain language
A sabotaged AI that behaves normally most of the time but performs a specific malicious action when it detects a hidden signal. Like slipping instructions into a recruitment AI that always approves a particular candidate when they apply.
Why this matters
Backdoor attacks represent a critical supply chain risk for organisations using pre-trained models or third-party AI components. Your governance framework must include verification testing and audit procedures to detect embedded backdoors before deploying external models in sensitive decisions.
Relevance
ImplementationBackdoor attacks require specific technical detection methods during model evaluation and testing before deployment, making this an implementation-level control.
Putting backdoor attack into practice in your organisation?
Ready to transform your AI strategy?
Partner with Australia's AI strategy and governance specialists. From adoption roadmaps to ISO 42001 audit readiness.