Back to Posts

Risk management and cognitive technologies

Cognitive technologies are changing the way we conduct risk analysis. The traditional tools of risk analysis just scratch the surface and greater insight is not possible due some limitations in terms of human capabilities,  which cognitive technologies can easily overcome. For example, any variance in the data can be studied under traditional risk management. However not all the variance can be flagged as risk. Sometimes those variances (false positives) can occur be due to some factors that can be ignored. However, this can take significant amount of time and effort. Cognitive technologies can learn from these special circumstances and next time it won’t flag those variances and can give better results thus saving significant amount time.

In case of some risk analysis if you have to select your best algorithm for your risk model, a significant amount of time is spent in permutation and combination of different algorithms so that you can have best model that can give you best result. Cognitive technologies can save time by automatically selecting the best algorithm that fits your business problem. Not only that but also for Cyber risk, compliance risk, and in some cases risk of human errors can also be tackled by cognitive technologies. With the help of big data, cognitive technologies can help build a strong risk protection system.

Here are some of the cognitive technologies that can help an organization tremendously in risk analysis and risk mitigation.

SparkCognition- Sparkpredict: It uses AI to optimize industrial maintenance. It uses predictive machine learning to analyse live streaming of sensor data and generates alarm when any deviation is identified. SparkPredict analyses sensor data, and usage of Machine learning for insight to predict possible machine failure. This saves loss in terms of downtime, and costly schedule maintenance

DeepArmor: It gives protection against zero-day malware, weaponized scripts, macros, and in-memory attacks by using power of big data and patented machine learning algorithms.

DeepNLP: It focuses on risk of human error in dealing with unstructured data. It uses Machine learning to create structure from natural language. Documents like contracts, reports, emails, pdfs, word, plain texts etc can be organized into logical groupings and then it can be searched in a logical and faster way. It uses deep-learning to create multiple overlays to look at same collection through different lenses, topics etc so that it can give better insight

Darwin: It accelerates analysis process by automating building and deployment of models. You need to select the kind of problem you are trying to solve, upload the dataset, and you need to specify the training parameters. Darwin uses multiple evolutionary algorithm and deep learning methods to automatically clean the data and iterate thousands of possible model architecture to find the optimum solution. It includes Python STK and REST API

Microsoft cognitive services (Anomaly Detector): It detects any anomaly in business by looking at time series data. It is somewhat similar to Darwin, where Anomaly detector looks at the time series data and automatically selects best algorithms

IBM Watson OpenSacle: Gives accurate view of AI system, monitors AI performance and fine tunes it. For example, if an insurance company is using AI to process claims, the system suggests to reject an application. The explain-ability feature gets detailed answer why claim was rejected if the customer or the regulator asks; also the bias feature automatically detects bias and mitigates it

IBM Watson Discovery: Extracts valuable contents from unstructured data. Built in NLP capabilities that analyses sentiments, entities, and others

IBM Watson Compare and comply: It extracts data from contracts and legal documents so that the non-compliance risk can be analysed and mitigated. It can analyse variety of unstructured data like PDFs, tables, or other unstructured data and it extracts data to compare any risk

Cisco Cognitive Threat Analytics: It automatically analyses more than 10 billion web requests daily. It focuses its attention on malicious activity that may have bypassed security controls and is using web-based communications.

Here is the detailed comparison:

  • SparkCognition
    • Sparkpredict
      • What it does? Asset protection and optimization
      • What risk it tackles? Use AI to optimize industrial maintenance. Uses predictive machine learning to analyse live streaming of sensor data and generates alarm when any deviation is identified
      • How it works? SparkPredict analyses sensor data, and usage of Machine learning for insight to predict possible machine failure. This saves loss in terms of downtime, and costly schedule maintenance
    • DeepArmor
      • What it does? Protection against zero-day malware, weaponized scripts, macros, and in-memory attacks
      • What risk it tackles? Prevent cyber attack
      • How it works? By using power of big data and patented machine learning algorithms
    • DeepNLP
      • What it does? Streamlines unstructured data
      • What risk it tackles? Risk of human error in dealing with unstructured data
      • How it works? Use Machine learning to create structure from natural language. Documents like contracts, reports, emails, PDFs, word, plain texts etc can be organized into logical groupings and then it can be searched in a logical and faster way. It uses deep-learning to create multiple overlays to look at same collection through different lenses, topics etc so that it can give better insight
    • Darwin
      • What it does? Automates building and deployment of models
      • What risk it tackles? Can be applied to predict to mitigate various risks
      • How it works? Accelerates data science at scale by automating building and deployment of models. You need to select the kind of problem you are trying to solve, upload the data set, and you need to specify the training parameters. Darwin uses multiple evolutionary algorithm and deep learning methods to automatically clean the data and iterate thousands of possible model architecture to find the optimum solution. It includes Python STK and REST API
  • Microsoft cognitive services
    • Anomaly Detector
      • What it does? It detects any anomaly in business by looking at time series data
      • What risk it tackles? Can be applied for various operational risk/ financial risk / hazard risk
      • How it works? It is somewhat similar to Darwin, where Anomaly detector looks at the time series data and automatically selects best algorithms
  • IBM
    • Watson OpenSacle
      • What it does? Gives accurate view of AI system, monitors AI performance and fine tunes it
      • What risk it tackles? AI Black box explanation problem, Bias feature
      • How it works? For example if an insurance company is using AI to process claims, the system suggests to reject an application. The explain-ability feature gets detailed answer why claim was rejected if the customer or the regulator asks. Also the bias feature automatically detects bias and mitigates it
    • Watson Discovery
      • What it does? Extracts valuable contents from unstructured data
      • What risk it tackles? Risk of human error in dealing with unstructured data
      • How it works? Built in NLP capabilities that analyses sentiments, entities, and others. It gives precise analysis of unstructured data for better decision
    • Watson Compare and comply
      • What it does? It extracts data from contracts and legal documents so that the non compliance risk can be analyzed and mitigated
      • What risk it tackles? Compliance or other contract related risks
      • How it works? It can analyze variety of unstructured data like PDFs, tables, or other unstructured data and it extracts data to compare any risk
  • Cisco
    • Cognitive Threat Analytics
      • What it does? Automatically analyzes more than 10 billion web requests daily. It zeroes in on malicious activity that has bypassed security controls and is using web-based communications. This includes standard, encrypted, and anonymous channels that can be used to attack organization
      • What risk it tackles? Cyber risk
      • How it works? Using machine learning and a statistical modeling of networks, Cognitive Threat Analytics creates a baseline of normal activity and identifies anomalous traffic occurring within network. It analyzes device behavior and web traffic to pinpoint command-and-control communications and data ex filtration.

 

Share this post

Back to Posts