Security of AI-Powered Systems: Threat Intelligence on the Edge

As the systems driven by artificial intelligence (AI) become increasingly integrated across various applications, ensuring their security is paramount to mitigate potential risks and vulnerabilities. Threat intelligence plays a pivotal role in identifying, analysing, and mitigating cybersecurity threats, particularly those targeting AI systems. Securing these systems entails safeguarding data integrity and confidentiality while ensuring the reliability and trustworthiness of AI algorithms and models. Challenges in AI-powered systems encompass adversarial attacks, model poisoning, data privacy concerns, as well as issues of bias and fairness. Threat intelligence offers valuable insights into emerging cyber threats, empowering organizations to proactively identify and address potential risks. Leveraging AI and machine learning techniques, threat intelligence platforms analyse extensive datasets to detect patterns indicative of cyber threats against AI systems. Sharing threat intelligence among organizations enhances collective defence capabilities and facilitates proactive threat mitigation strategies. This paper aims to propose an AI-powered systems cyber risk analysis approach, encompassing machine learning techniques. To achieve that, an intrusion detection simulation was done by generating a Python script with synthetic data. Two datasets were generated - a training dataset with 10,000 records and a test dataset with 400 records. Machine learning techniques are applied to predict what the most common security issues are.