Are there any good algorithms which I could use and combine to further improve cyber security? What would be the future? is AI a good solution? how can we contribute to this? help please, thank you!
I would start with an overview on how attackers use AI or machine learning to attack IT systems. AI is only in a last instance something like algorithms! Look to innovative design concepts to build robust IT which is intrinsic secure for cyber attacks.
There are several problems within the Cyber security field can be solved by using A.I techniques. Such as:
1- Network Intrusion Detection System (IDS):
how to classify the network behavior of the user into normal behavior or attack?
How to enhance the classification accuracy of the anomaly detection which is a type of IDS. In this case, you can apply any machine learning models (i.e., classifiers) such as Artificial Neural Network (ANN), Support Vector Machine (SVM), Decision Tree (DT), or Naive Bayesian Classifier (NBC).
The recent research papers in this field try to enhance the classifiers by using metaheuristics, for example, an algorithm used to train ANN (i.e., finding the weights of the nodes), or finding the optimal parameters for SVM... etc..
One more research topic where A.I can be used to enhance IDS, by selecting the most relevant features in the IDS dataset (i.e., NSL or KDD dataset ). This problem is called Feature Selection Problem. The feature selection simply depends on one important question which is , How to select the minimum number of features which enhance the classification accuracy ?
i.e., Min (No. of Features) --> Max (Classification Accuracy).
2- Email or SMS spam filtering ..
how to classify the emails or the sms text into spam and not spam ? In this problem, A.I or machine learning models can be applied exactly same as the previous problem.
3- Malware detection problem. Same as the previous points..
4- Intelligent Anti virus softwares.
Generally the AI can be used as an effective set of tools in cyber-security including IDS, automatic behavior analysis, human-computer interaction patterns analysis, and etc.
Would that be a good research venture? AI enabled IDS? Which algorithms could possibly be a good use for this? I would love to do research with regards to this topic
Hello Francis,
One really good thing to research on your venture would be how AI could implement cryptographic methods to improve cyber security as well as, securing the network. AI has to be a blended approach. With advances in technology, speed of computations and attackers utilizing GPU our cryptographic methods may soon be extinct. Another poster mentioned behavior analysis which is another really good point. I am going to add a link here where you can read more about AI and cryptography:
Blackledge, J., Bezobrazov , S., & Tobin, P. (2015, July). Cryptography using Artificial Intelligence. Retrieved from Research Gate:
I would further your research here. One point I would like to make is that AI may be a great solution to prevent intrusions although, what happens if attackers are utilizing AI? Of course, limiting an attacker to a specific segment helps but on the other hand, if they get in what prevents the attacker from gaining access to proprietary information? Something to think about.
Jason Caputo awesome stuff!! please keep it coming! ok so the focus with be AI doing the cryptography to increase them defense mechanism.
Francis Jesmar Perez Montalbo network defense is a multi-layer approach. There are many factors that come into play. Such as securing the network environment from intruders by using many layers. Your research should focus on the multi-layered approach with AI as well as how using AI to enforce cryptography should an attacker gain access to a network environment. Does that clarify for you?
In order for AI to truly be successful in any environment it would have to be a multi-layered approach just as we do today but better.
Jason Caputo wow! thank you for that sir! Just hoping if you can recommend good materials and references for it. I would love to study those, thank you for those.
In addition to the excellent answers already provided, you can look at more specific areas within IDS. For example User Behavior Analytics aims at providing more focused flavor of anomaly detection by narrowing observations to users,devices etc. This aids in overcoming the big data challenge of analyzing entire network activities while detecting advanced threats like APTs and insider attacks. Similarly, there are open research areas within cloud security, IoT security etc.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence