Between data and defense | KPMG | CH

Between data and defense. How technology can help protect us.

Between data and defense

Ann Johnson, Vice President of Worldwide Cyber Security at Microsoft, shares her insights into the future of ML and AI.

1000

Related content

Ann Johnson

Ann Johnson, Vice President of Worldwide Cyber Security at Microsoft

Machine learning (ML) and artificial intelligence (AI) have become an integral part of our life. Which developments will impact us most over the next three to five years?

Developments in this area include new medical research-based data that can deliver improvements in healthcare. Patients will ultimately be able to benefit from medical research that uses ML and AI to predict medical conditions, enhance patient care, and even develop new and better drugs. A further development will be our growing reliance on machines to carry out tasks that are typically requested of humans. Enhanced voice input- based machine assistance is becoming particularly popular. To date, people use voice-assistant software such as Microsoft’s Cortana to carry out simple commands such as finding facts, files, places, other information, setting timers and reminders, etc. Over time, these technologies will become more sophisticated, applying ML and AI to understand the context behind the original command and even to automatically carry out other, related tasks. And last but not least, enhanced and timely detection of cyber security attacks. Cyber threat intelligence derived from the application of ML and analytics will continue to augment cyber security analysts’ efforts to protect endpoints, better detect attacks and accelerate responses. ML and data analytics are already used extensively to detect nation-state attacks. For instance, the Microsoft Intelligent Security Graph, based on Microsoft Azure Machine Learning technology, collects trillions of data points from billions of computing endpoints to provide real- time insights that are of great use in this field. The Office 365 Advanced Threat Protection product alone processes 6 billion emails each day.

Do you see current/future risks to the privacy of people (data subjects) as a consequence of using ML and AI? What do you consider to be effective means to mitigate such risks?

It is possible that such risks exist. ML algorithms that don’t have security and privacy built-in by design may accidentally or inadvertently expose personal data. For example, we wouldn’t want voice-assistant software on our device to share Personally Identifiable Info or PII for anyone to see or hear. One way to mitigate risks would be to ensure all private data, including PII data, is encrypted. And that ML and AI algorithms transact on encrypted data. Fortunately, advancements in the field of mathematics are already enabling this. The medical industry, in particular, is adopting this approach to safeguard patient and hospital data and securely analyze data at scale.

Do you see current or future cyber risks for individuals, organizations and states as a consequence of using ML and AI? If so, how should these risks be mitigated?

Organizations are responsible for maintaining privacy for a wide variety of data related to the business itself, its employees, customers and partners. If the application of ML or AI compromised these data, the company could suffer reputational, financial or other damage. Meanwhile, nation states similarly need to maintain the integrity and privacy of all types of data covering a wide variety of topics such as industrial, financial and people. It’s imperative that vendors of ML and AI technologies fully disclose to state entities how they ensure the privacy of their data. The same goes for individual’s privacy, one way to mitigate risk would be to ensure all private data, including PII data, is encrypted and that ML and AI algorithms can process these data.

Which do you consider the most promising use cases for ML and AI for cyber defense? Are you working on such use cases and, if so, what are the potential benefits?

Among the most promising use cases in this regard are the application of the NIST Cyber Security Framework to protect, detect and respond to cyber threats. Protection focuses on applying security controls to minimize the impact of a potential cyber security event. ML-based solutions help prevent compromised identity, secure applications and data, expand device controls (including data encryption), and safeguard infrastructure. Benefits can also be had from using security solutions that use ML and AI to help detect advanced cyber attacks. The solutions seek out signs of malicious behavior, anomalies, and other noteworthy cyber security events. Meanwhile, response focuses on being able to lessen the impact of a potential cyber security event. Organizations should apply a multi-pronged approach of training people, developing and adhering to security best practice processes and implementing cyber security technology. Microsoft offers clients an unparalleled body of threat intelligence that is created from analysis across our customer base, including emails, URLs and endpoints. Our security solutions apply a combination of artificial and human intelligence to provide visibility into suspicious and malicious activity – enabling customers to respond and mitigate cyber threats swiftly.

Do you have any final comments on this subject that may be of interest to our readers?

ML and AI have the potential to drive improvements and help with a variety of use cases for consumers, enterprises and government. However, these technologies will only be adopted and used successfully at scale if they transact on trusted data and do not in any way compromise the security and privacy of personal data. Then, we can see huge benefits across a number of vital fields.

© 2017 KPMG Holding AG is a member of the KPMG network of independent firms affiliated with KPMG International Cooperative (“KPMG International”), a Swiss legal entity. All rights reserved.

Connect with us

 

Request for proposal

 

Submit