cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Kyaw_Myo_Oo
Contributor III

Secure AI System Development Guidelines

Dear Community Members,

 

The UK National Cyber Security Centre, the US Cybersecurity and Infrastructure Security Agency (CISA) and similar organizations from 16 other countries have published guidelines for secure AI system development. The guidelines address four stages in the development lifecycle: secure design, secure development, secure deployment, and secure operation and maintenance.

Secure AI System Development Guidelines.

 

https://www.ncsc.gov.uk/files/Guidelines-for-secure-AI-system-development.pdf?is=2e17210a000df3b2371...

 

 

Kyaw Myo Oo
Manager , CB BANK PCL
CCIE #58769 | PCNSE | CCSE | CISSP | PMP
2 Replies
Soniya-01
Newcomer II

Creating a secure AI system is similar to building a superhero with a strong shield. Here are some entertaining and simple rules to ensure that your AI superhero remains safe and sound: Armor Up Your AI: Just like a superhero needs a solid suit, your AI system needs strong security measures. Put on firewalls, encryption, and all the cool cyber armor to protect against bad guys. Train Your AI Hero: Superheroes do not become great overnight. Train your AI with good data so it knows the difference between a villain and a friend. Be careful not to teach it the wrong lessons! Guard the Secret Identity: Every superhero has a secret identity.

Caute_cautim
Community Champion

@Soniya-01    You obviously watch Marvel movies too - super heroes.

 

You might find the following reference for OWASP useful in terms of modelling the threats facing AI Models:

 

https://owasp.org/www-project-ai-security-and-privacy-guide/owaspaiexchange.html

 

They provide a very good model for identifying the threats and is loaded with references to standards, which I hope many will find useful.  This is not a trivial job, more than application testing, requires involvement of ongoing security & privacy by design principles and practices.

 

Regards

 

Caute_Cautim