The learning rate is defined as a hyper-parameter that controls the weights of a neural network with respect to gradient loss. It defines how quickly the neural network updates the concepts it has learned. How does SilverSky use learning rate and what is the role of artificial intelligence (AI) in it? Let’s find out.
SilverSky’s first exposure to the learning rate discussion occurred during its collaboration with the Defense Advanced Research Projects Agency (DARPA). Our Chief Artificial Intelligence Scientist, Rich Heimann, provides more insight into this process in his book “Doing AI” explaining how using AI is both complicated and difficult to create, but when it is applied correctly, it is very powerful. The basic concept of the learning rate problem is that today’s cyber adversaries (bad guys) are different from the past. Villains have become more sophisticated, specialized, and organized. Based on their improvements, they developed sophisticated teams, each responsible for different aspects of attacking a network, application, etc. They are very similar in this respect to the way most companies run their information technology (IT) organization by having specialists for the network and another team for the applications etc. The bad guys even went a step further by adding teams that focus on extortion, once the intrusion is successful. The bad guys aren’t bored teenagers anymore, but actually start-up companies that may or may not be sponsored by other countries that fund them, just like Silicon Valley venture capitalists fund their security startups. The whole ecosystem is well funded, highly specialized, globally integrated and highly sophisticated.
The result is that the defensive cyber resilience ecosystem (the good guys) has failed to keep pace with the offensive ecosystem of the hacker community. The resources used by a company’s security operations organizations, the level of training of operations teams, and the processes those teams follow have lagged behind the advancement of the bad guys. If this were to be illustrated, it would reflect that the bad guys are getting new tactics to hack into a system at an alarming rate compared to the good guys. This presents a scenario that, as each day passes, the gap between the bad guys’ ability to penetrate and the good guys’ ability to defend themselves widens.
As a member of the good guys, we need to learn new ways to keep up with this rapidly changing skill set that the bag villains present. We need to keep up with the levels of automated processing, machine learning, advanced intelligence, and a well-trained workforce that the bad guys use to attack our networks and systems every day. Because of this advanced level of automation, the cost of entry for bad guys is so low that no business is safe from the crosshairs of small businesses, mid-sized businesses, and large enterprises. …everyone is fair game these days.
Things good guys need to focus on:
- Training – our teams need to focus on learning and improving every day
- Processes – patches, system upgrades and human internal communication processes, our teams must commit to “not dropping a drop”
- Tools – ensuring tools can learn from data, continually improving with the ability to create detections on the more obscure indicators to try and close the gap with the bad guys. This is where AI can help bridge the gap
- Coverage – providing both broad and deep detection capabilities so that even small anomalies are assessed to ensure they are not part of a larger attack
- Look to partner with others to ensure you are getting the most out of your security budget. None of us can do it alone
Remember that security tools or even vendors shouldn’t be analyzed solely on price or old-fashioned minimal thought processes. You need to analyze the depth of capabilities and results:
- Do these vendor tools and capabilities have the ability to protect my organization?
- Can these tools keep pace with advanced villain sophistication and does the vendor have the drive to learn every day and be the best?
- Does my vendor want to be an extension of my IT team and not just see us as a “customer”?
The best way to achieve the above goals is to make sure you start with a platform with sophisticated advanced analytics that uses AI and machine learning (ML) techniques designed specifically to keep pace. quick learning from the bad guys. Additionally, you and/or your security vendor should have a highly trained operations team to respond immediately and thoroughly when even the slightest indicator of compromise is detected. Look for these characteristics when choosing your tools and supplier. Start by acknowledging that villains learn at a rapid pace. Then look for a plan to keep up with that accelerated pace of learning, and the tools and people who can execute that plan.
SilverSky is able and ready to provide you with a comprehensive assessment and service to protect your organization’s assets. Get a demo of our Lightning services and see how we can help you.
*** This is a syndicated blog from the Security Bloggers Network of SilverSky written by joe-neri. Read the original post at: https://www.silversky.com/blog/what-is-rate-of-learning/