Neural Networks and Artificial Intelligence History

The Use of Neural Networks In Cybersecurity Ian Wolff September 19, 2018

History of Artificial Intelligence(AI) An idea that goes back to early Chinese and Greek cultures with Automatons. Alan Turing, 1950 paper on Computing Machinery and Intelligence , suggested that humans use available information as well as reason in order to solve problems and make decisions, so why can t machines do the same thing? AI was officially coined in 1956, during a conference at Dartmouth College, by Allen Newell, Cliff Shaw, and Herbert Simon. Program called Logic Theorist funded by the Research and Development Corporation (RAND). The scientific community had an off again, on again love of the AI field until 1997. IBM s Deep Blue beat Chess grandmaster Garry Kasparov. IBM s Watson won Jeopardy in 2011.

History of Artificial Intelligence(AI) Cont.

The Different Fields within Artificial Intelligence(AI)

Overview of Artificial Intelligence A major thrust of AI is in the development of computer functions associated with human intelligence, such as reasoning, learning, and problem solving. Artificial intelligence is a science and technology based on disciplines such as Computer Science, Biology, Psychology, Linguistics, Mathematics, and Engineering. Artificial intelligence, sometimes called machine intelligence, is intelligence demonstrated by machines, in contrast to the natural intelligence displayed by humans and other animals. Unsupervised vs Supervised Supervised learning are datasets that we have a reasonable idea of what the output should look like. Contains labels Unsupervised learning allows for a function to classify and group data without knowing what the structure should look like. Learns similarities without names, can identify what s different. Reinforcement Learning Goal oriented algorithms How to obtain max number of points in a videogame? DeepMind uses this method to create a neural network that functions nearly the same a human s short term memory.

What are Neural Networks? Inspired by the animal brain s processing power and composition. Each artificial neuron cab transmit a number between 0 and 1. Leverages computational inputs, Sigmoids (Logistic Curve) and weights. Comes in different flavors. Also referred to in some cases as Deep Learning, or Deep Learning Networks.

What are Neural Networks, cont.

Brief Example of Neural Networks https://www.youtube.com/watch?v=aircAruvnKk But what *is* a Neural Network? | Deep learning, chapter 1 By 3BLUE1BROWN

Applications of Neural Networks within Cybersecurity Network based Misuse and Anomaly Detection HIDS and NIDS Detection and Response MIT Research using neural Network IDS in 2011 30 keywords used to detect attacks Cat, uuedecode, root shell commands 17 of 20 attacks detected with 1 false positive during attacks using shell code commands. 68 of 73 attacks detected with 4 false positives during attacks that executed C source code. Endpoint Detection and Response Cylance 2.7 Quadrillion turns or combinations used to detect anomalous behavior 99% True Positive Rate (Their words, not mine) What AV should have been 5 years ago

Applications of Neural Networks within Cybersecurity cont. Threat Intelligence Platforms Leverage deep learning and Machine Learning to categorize and report on: Pattern recognition Natural Language Processing Adds structure to unstructured text (slang, jargon, multi-language support) Predictive Analytics

Applications of Neural Networks within Cybersecurity, cont. SignalSciences Next Gen WAF and Runtime Application Self Protection (RASP) Leverages neural networks to detect attack patterns and create response signatures. Also able to analyze traffic and notify customers on anomalous activity.

The Downside to Neural Networks and AI within Cybersecurity Explainability Why did this detection and response event occur? Complex algorithms are difficult to explain Supervised Datasets Currently, limited to human created datasets, what did we miss? Unfinished Products Hitting the Market The current hype cycle means a released product that may not have been completely vetted. Too much training data and not enough real-world data used.

Wrapping up Investing in AI is a good thing for our industry. Almost certainly behind the bad guys. As always, properly vet your products to ensure they fit you environment, with an emphasis on what s under the hood. Neural Networks and Deep Learning will allow us to leverage all manner of data, with the hopes of being sent actionable high fidelity data.

References https://towardsdatascience.com/supervised-vs-unsupervised-learning-14f68e32ea8d https://www.youtube.com/watch?v=aircAruvnKk https://en.wikipedia.org/wiki/Artificial_neural_network http://pages.cs.wisc.edu/~bolo/shipyard/neural/local.html https://www.sans.org/reading-room/whitepapers/detection/application-neural-networks-intrusion- detection-336 https://www.technologyreview.com/s/611860/ai-for-cybersecurity-is-a-hot-new-thing-and-a-dangerous- gamble/ https://www.recordedfuture.com/machine-learning-cybersecurity-applications/ https://www.signalsciences.com/announcements/network-learning-exchange-power-rules-announcement/ https://en.wikipedia.org/wiki/DeepMind https://skymind.ai/wiki/deep-reinforcement-learning http://sitn.hms.harvard.edu/flash/2017/history-artificial-intelligence/ https://www.livescience.com/49007-history-of-artificial-intelligence.html https://becominghuman.ai/what-is-artificial-intelligence-ai-4bde325e5462 https://www.forbes.com/sites/forbestechcouncil/2018/07/11/machine-learning-vs-artificial-intelligence- how-are-they-different/#312bf6703521 https://threatvector.cylance.com/en_us/home/how-traditional-antivirus-works.html