人工智能全生命周期安全风险综合评估方法

Research on Comprehensive Assessment Method of AI Lifecycle Security Risk

人工智能技术在不断发展的同时,其在各行业中的使用也越来越广泛。但是人工智能技术在为人们的生产生活带来便利的同时,也引入了新的安全风险。为了应对潜在的安全隐患,做到事前预防,保障人工智能的安全,提出了一种基于信息熵和马尔可夫链的人工智能全生命周期安全风险评估方法。一方面,该方法能够在对人工智能进行整体性的安全风险等级评估的同时,保留对各类安全风险的评估;另一方面,基于马尔可夫链形成各个风险类在稳定状态下的转移矩阵和概率分布,提供了一种面向全局的人工智能安全风险的展现方法。

With the continuous development of AI technology,its application in various industries is becoming more and more extensive.However,while AI technology brings convenience to hunman’s production and life,it also introduces new security risks.In order to deal with potential security risks,achieve prior prevention,and ensure the security of AI,this paper proposes a lifecycle security risk assessment method of AI based on information entropy and Markov chain.On the one hand,the method can retain the assessment of various security risks while performing an overall security risk level assessment of AI;On the other hand,based on Markov chains to form transition matrix and probability distribution of each risk class in steady state,it provides a global oriented method for presenting AI security risks.