Mobile devices are popular with hackers because they’re designed for quick responses based on minimal contextual information. Verizon’s 2020 Data Breach Investigations Report (DBIR) found that hackers are succeeding with integrated email, SMS and link-based attacks across social media aimed at stealing passwords and privileged access credentials. And with a growing number of breaches originating on mobile devices according to Verizon’s Mobile Security Index 2020, combined with 83% of all social media visits in the United States are on mobile devices according to Merkle’s Digital Marketing Report Q4 2019, applying machine learning to harden mobile threat defense deserves to be on any CISOs’ priority list today.
How Machine Learning Is Helping To Thwart Phishing Attacks
Google’s use of machine learning to thwart the skyrocketing number of phishing attacks occurring during the Covid-19 pandemic provides insights into the scale of these threats. On a typical day, G-Mail blocks 100 million phishing emails. During a typical week in April of this year, Google’s G-Mail Security team saw 18M daily malware and phishing emails related to Covid-19. Google’s machine learning models are evolving to understand and filter phishing threats, successfully blocking more than 99.9% of spam, phishing and malware from reaching G-Mail users. Microsoft thwarts billions of phishing attempts a year on Office365 alone by relying on heuristics, detonation and machine learning strengthened by Microsoft Threat Protection Services.
42% of the U.S. labor force is now working from home, according to a recent study by the Stanford Institute for Economic Policy Research (SIEPR). The majority of those working from home are in professional, technical and managerial roles who rely on multiple mobile devices to get their work done. The proliferating number of threat surfaces all businesses have to contend with today is the perfect use case for thwarting phishing attempts at scale.
What’s needed is a machine learning engine capable of analyzing and interpreting system data in real-time to identify malicious behavior. Using supervised machine learning algorithms that factor in device detection, location, user behavior patterns and more to anticipate and thwart phishing attacks is what’s needed today. It’s a given that any machine learning engine and its supporting platform needs to be cloud-based, capable of scaling to analyze millions of data points. Building the cloud platform on high-performing computing clusters is a must-have, as is the ability to iterative machine learning models on the fly, in milliseconds, to keep learning new patterns of potential phishing breaches. The resulting architecture would be able to learn over time and reside on the device recursively. Protecting every endpoint if it’s connected to WiFi or a network or not is a key design goal that needs to be accomplished as well. MobileIron recently launched one of the most forward-thinking approaches to solving this challenge and its architecture is shown below:
Five Ways Machine Learning Can Thwart Phishing Attacks
The one point of failure machine learning-based anti-phishing apps continue to have is lack of adoption. CIOs and CISOs I’ve spoken with know there is a gap between endpoints secured and the total endpoint population. No one knows for sure how big that gap is because new mobile endpoints get added daily. The best solution to closing the gap is by enabling on-device machine learning protection. The following are five ways machine learning can thwart phishing attacks using an on-device approach:
2. Using machine learning to glean new insights out of the massive amount of data and organizations’ entire population of mobile devices creates a must-have. There are machine learning-based systems capable of scanning across an enterprise of connected endpoints today. What’s needed is an enterprise-level approach to seeing all devices, even those disconnected from the network.
3. Machine learning algorithms can help strengthen the security on every mobile device, making them suitable as employees’ IDs, alleviating the need for easily-hackable passwords. According to Verizon, stolen passwords cause 81% of data breaches and 86% of security leaders would do away with passwords, if they could, according to a recent IDG Research survey. Hardening endpoint security to the mobile device level needs to be part of any organizations’ Zero Trust Security initiative today. The good news is machine learning algorithms can thwart hacking attempts that get in the way making mobile devise employees’ IDs, streamlining system access to the resources they need to get work done while staying secure.
4. Keeping enterprise-wide cybersecurity efforts focused takes more than after-the-fact analytics and metrics; what’s needed is look-ahead predictive modeling based machine learning data captured at the device endpoint. The future of endpoint resiliency and cybersecurity needs to start at the device level. Capturing data at the device level in real-time and using it to train algorithms, combined with phishing URL lookup, and Zero Sign-On (ZSO) and a designed-in Zero Trust approach to security are essential for thwarting the increasingly sophisticated breach attempts happening today.
5. Cybersecurity strategies and the CISOs leading them will increasingly be evaluated on how well they anticipate and excel at compliance and threat deterrence, making machine learning indispensable to accomplishing these tasks. CISOs and their teams say compliance is another area of unknowns they need greater predictive, quantified insights into. No one wants to do a compliance or security audit manually today as the lack of staff due to stay-at-home orders makes it nearly impossible and no one wants to jeopardize employee’s health to get it done. CISOs and teams of security architects also need to put as many impediments in front of threat actors as possible to deter them, because the threat actor only has to be successful one time, while the CISO/security architect have to be correct 100% of the time. The answer is to combine real-time endpoint monitoring and machine learning to thwart threat actors while achieving greater compliance.
For machine learning to reach its full potential at blocking phishing attempts today and more advanced threats tomorrow, every device needs to have the ability to know if an email, text or SMS message, instant message, or social media post is a phishing attempt or not. Achieving this at the device level is possible today, as MobileIron’s recently announced cloud-based Mobile Threat Defense architecture illustrates. What’s needed is a further build-out of machine learning-based platforms that can adapt fast to new threats while protecting devices that are sporadically connected to a company’s network.
Machine learning has long been able to provide threat assessment scores as well. What’s needed today is greater insights into how risk scores relate to compliance. Also, there needs to be a greater focus on how machine learning, risk scores, IT infrastructure and the always-growing base of mobile devices can be audited. A key goal that needs to be achieved is having compliance actions and threat notifications performed on the device to shorten the “kill chain” and improve data loss prevention.