Madhav Chablani

'Data Security Council of India (DSCI) introduces 'Data Protection Seal' for Complaint Organizations which are found to be using data safely and following the basic standards for data privacy. The DPS will be on similar lines to the ISI mark which is evidence that a product is under the Bureau of Indian Standards. According to the CEO of DSCI Vinayak Godse: 'The data protection seal will provide some level of assurance about the application, website, or product, according to expectations of privacy, and whether it behaves responsibly'. The DSCI has been quite active in coming up with initiatives in the data privacy and data protection sector. It has recently started a data protection credential programme, known a DSCI-certified Data Protection Officer, which helps organisations in getting their DPOs trained as per the requirements under the DPDP Act 2023. The DPS will serve as an incentive for firms to comply faster and comprehensively with DPDP Act and subsequent rules, and notifications.' 'DSCI is a not-for-profit set up by the industry body - NASSCOM (National Association of Software and Service Companies - works in the areas of cybersecurity and privacy and engaged with governments, agencies, regulators, industry sectors, associations and think tanks for policy advocacy, capacity building and outreach activities'. https://lnkd.in/gGravBHU

1

AI RISK MEASUREMENT AI risk measurement involves assessing the potential risks associated with deploying artificial intelligence systems. Tools and Frameworks for AI Risk Measurement: 1. AI Fairness 360 (AIF360): - Developed by IBM, AIF360 is an open-source toolkit that helps detect and mitigate bias in machine learning models. 2. AI Explainability 360 (AIX360): - Toolkit that supports interpretability and explainability of machine learning models. - It includes algorithms, metrics, and tools to make AI models more transparent and understandable. 3. Ethical AI Frameworks: - Organizations like the IEEE and the European Union have developed ethical guidelines for AI development. 4. Model Governance Tools: - Tools like ModelOp Center help manage and govern AI models, ensuring compliance with regulatory requirements and internal policies. 5. Risk Assessment Frameworks: - The AI Risk Management Framework (AI RMF) from NIST (National Institute of Standards and Technology) offers guidelines for assessing and managing risks associated with AI. 6. Privacy Impact Assessment (PIA) Tools: - Tools such as TrustArc and OneTrust help conduct privacy impact assessments to identify and mitigate privacy risks in AI systems. - These tools ensure compliance with data protection regulations like GDPR and CCPA. 7. Adversarial Robustness Toolkits: - Libraries like IBM's Adversarial Robustness Toolbox (ART) provide tools to test and improve the robustness of machine learning models against adversarial attacks. 8. Bias and Fairness Assessment Tools: - Google's What-If Tool allows users to analyze machine learning models for fairness and interpretability. 9. Security Assessment Tools: - Tools like Microsoft's Counterfit provide a framework for assessing the security of AI systems. - They simulate various attack vectors to identify vulnerabilities. 10. Operational Risk Management Tools: - Platforms like SAS Risk Management and Palantir Foundry offer comprehensive risk management solutions that include AI and machine learning capabilities. Integrating AI Risk Measurement: 1. Continuous Monitoring: - Implement continuous monitoring to detect and respond to risks in real-time. - Use tools like Splunk and DataRobot's MLOps. 2. Regular Audits and Reviews: - Conduct regular audits and reviews of AI systems to ensure compliance with ethical, legal, and operational standards. - Engage third-party auditors to provide independent assessments. 3. Stakeholder Engagement: - Involve stakeholders, including customers, employees, and regulatory bodies, in the risk assessment process. - Gather feedback and address concerns to build trust and transparency. 4. Documentation and Reporting: - Provide transparent reporting to stakeholders on the measures taken to mitigate risks. #AISECURITY #riskmanagement

3

Very interesting multi-part question from Vijayan Packeriswamy in LinkedIn Learning's 'Artificial Intelligence for Cybersecurity' course 3 Questions on AI and Cyber Security: 1. How susceptible are AI systems to adversarial attacks?   2. How can we prevent data poisoning attacks on AI systems?   3. How can we ensure transparency and interpretability in AI models used for cybersecurity? I'm re-posting this to see what insights I can gather from all of my LinkedIn colleagues and connections currently working in Artificial Intelligence &/or Cybersecurity ... looking forward to any information and current thinking / practices happening in the industry today. Original questions posted here in the LinkedIn Learning Course: https://lnkd.in/dGYnvmPQ #artificialintelligence #ai #cybersecurity #security #vulnerabilities #datapoisoning #aimodels #dangerdanger

1