Active Intrusion Prediction and Response for Computing Devices on the Internet

加拿大以及世界各地的计算基础设施正越来越多地受到入侵者的攻击，这些入侵者试图窃取安全、商业和隐私关键信息。这些入侵者通常得到犯罪企业或民族国家行为者的支持，在技术上非常复杂，拥有大量资源。2015年，英国保险公司劳合社估计，全球因网络犯罪造成的损失达5000亿美元。到2019年，这一数字预计将翻两番，达到2万亿美元。对抗这一日益恶化的危机的关键武器是检测网络入侵并部署对策的工程解决方案。

Computing infrastructure in Canada as well as across the world are increasingly under attack from intruders attempting to steal information that is safety, business and privacy critical. Such intruders are often backed by criminal enterprises or nation-state actors and are technically very sophisticated with plenty of resources. In 2015, British insurance company Lloyd's estimated the world-wide loss due to cyber crime USD $500 billion. By 2019, this is expected to quadruple to $2 trillion. A crucial weapon in the fight against this worsening crisis is an engineering solution to detect network intrusions and deploy countermeasures. Current intrusion detection systems work by monitoring network activity and looking for activity patterns that match known attack steps. Often, this results in detection after the attack has taken place and response is limited to recovery operations instead of thwarting an ongoing attack. This proposal aims to develop innovative solutions for intrusion detection and response that far surpass the state-of-the-art in following ways: a) generate many possible attack scenarios when an initial stage of an intrusion is detected, b) deploy additional monitoring tools that are specifically designed to gather evidence to confirm or deny these attack scenarios, c) plan defensive measures that can effectively counter the most likely scenarios and d) deploy these defensive measures in a way that is is least disruptive while continually evaluating the situation as above. Long term vision is to develop intelligent intrusion response systems that are aware of the security situation within the next decade. The short term objectives in this proposal that are designed to achieve the above goals include developing a novel way to represent intrusion events that can capture high-level intent as well as low-level network activity, recognizing combinations of intrusion events that are part of a larger attack and deploying these tools within a framework that allows building situational awareness. In the long term, this proposal aims to develop algorithms that can predict impending penetrations based on identification of ongoing attacks and develop countermeasures that can thwart these predicted attacks. A highlight of the proposed research is the development of algorithms that can learn from past attacks as well as setup decoy targets and dynamically change the network paths to divert attacks away from critical systems. The latter will also allow gathering crucial data about these attacks that will be used to design better detection tools in the future. Information security is becoming a critical need for the foreseeable future and innovative approaches as well as highly trained professionals are needed to meet the looming challenges. This program will train the engineers with necessary background as well as with hands-on experience, who will be able to fill this critical and growing need.