HACK-O-HIRE

Problem Statement

Analyze the current and historical data which is available publicly and predict future Threats or Cyber Attacks and impact on any firm using web scrapping. Example: systems going down during Black Friday, Data Leakage Thefts, location impact on sites due to some riots etc.

Technology

• Python
• SQL
• Tableau
• Any database technology
• Excel, etc

Other Considerations

Along with Open data available on various websites, good to correlate with other information as well like Holidays, Public events, location etc.

Data

(Below are just examples) 

https://www.itgovernance.co.uk/blog/list-of-data-breaches-and-cyber-attacks-in-2023

https://cyberheatmap.isi.jhu.edu/

A group of faculty and students at the Johns Hopkins University have come together to develop this project and the resulting Cyber Attack Predictive Index (CAPI) described on this website. The predictive analysis is based on identifying common factors in select cyber-attacks among nation-states over the past 15 years.

Design Considerations

The solution should not be rigid with formatting of data, should be able to scale, utilize multiple sources and should be flexible. Tableau/Reporting solution should filter across the data sets and should give a better user experience.

Good to have

A dashboard correctly forecasting based on data publicly available with right filters. If it’s a known threat add Mitigation technique available online.

Advanced (Icing on the cake)

Use of ML algorithms to learn patterns and relationships allowing the code to make predictions or decisions without being explicitly trained for each scenario.

Benefits

• This will help us take right decisions and ensure better controls if we know potential threats/attacks.
• This will make businesses more secure and future ready.

Mentors

• Kaustubh Phophaliya
• Vignesh Kumar

Problem Statement

Create a Digital Box capability that

• Stores all customer communications
• Provides two-way interactive capabilities for complaints/queries
• Facilitates secure upload/download of documents at customer level

Technology

• ETL
• API
• Database (SQL/No SQL)  to store and retrieve the data.
• Cloud technologies, AWS preferable
• Open-source Encryption/decryption algorithm
• QlikView MI
• AI

Other Considerations

• (Nice to have features) Analytics and MI can help us identify the most urgent and negative complaints, prioritize them and respond accordingly. Text summarization can help us categorize the complaints from different sources and generate concise and accurate reports. Additionally, text classification can help us sort and filter the complaints based on their type, severity, product or service
• Digital Boxshall work as common repository and make data available while keeping Data Privacy intact.

Data

• XXX Customers active on Mobile App
• XXX different customer interactive channels

Design Considerations

• There are multiple digital channels to communicate/exchange document based on different product holdings. eg. SMS, In-app Notification, email etc.
• Each channel may have different ways to store logs.
• Store all documents , communications, complaints and resolution exchanged with customer
• Audit logs of each Customer Communication

Benefits

• A Digital Box helps colleagues see all available data/documents/queries/complaints for a customer to elevate customer experience and satisfaction index.
• Simplify communication with customer across products.
• Improved complaints handling
• Ease of handling documents

Mentors

• Vikas Bohara
• Aarti Jain
• Abhishek Neema

Problem Statement

• Any transaction settlements are processed based on the detailed trade level data which is sourced from various systems, further transformed and standardized.
• These trade feeds may have millions of records each day capturing all the trading activity and there are edge cases of inflated/missing prices, bad data, fat finger etc. which result in erroneous calculations and payments.

Develop an Anomaly Detection Framework that will help identify the potential issues and irregularities in the data when compared with the regular submissions.

Technology

• ETL
• Python
• Database (SQL/No SQL)  to store and retrieve the data.
• Open-source Encryption/decryption algorithm
• ML
• Tableau

Other Considerations

• ML serves as an efficient tool to read large data sets and formulate patterns.
• These generic frameworks can be deployed across the banks to learn using the massive data

Data Example

• XXX Feeds
• XXX Trades

Design Considerations

• Easy ways to load and scan through the data
• Build Similar patterns
• Feature Engineering
• Model Selection & Evaluation

 Benefits

• Accuracy of the Payments.
• Cost Savings owing to reduction in manual effort.
• Self-Learning Model which can be integrated across various functions within markets.

Mentors

• Pooja Choubey
• Mayur Kotak
• Bimal Mohapatra
• Akhil Vira

Problem Statement

Email classification based on the content.

Multiple emails from customers/clients that are dealt with different teams based on the context. This solution should enable auto-classification of emails based on the context, so the same can be routed to best suited team for further processing

Technology

• Python
• Anaconda platform

Other Considerations

The data selected should represent sufficient variation to be able to demonstrate classification clearly. Expectation from participants will be to present overall solution with clear focus on characteristics of data and holistic nature of the implementation.

Data

For solving this problem, participants can decide to leverage data available on public forums like Kaggle (preferably from finance domain). But the model should be easy to configure/retrain for similar topics.

Design Considerations

This model should be easy to deploy to execute either as batch or real time.

Focus should also be on making it efficient from resource consumption standpoint and something that can be hosted as containers.

Benefits

Auto-email classification will enable significant reduction in manual efforts

Mentors

• Dhivya Sekar
• Vijayshankar Subramanian
• Rahul Khare
• Shishir Shahi