Encrypting everything you store in the cloud is safe but expensive. Storing everything as plain text is cheap but risky. Confimizer uses AI to decide which data really needs encryption, saving money while keeping sensitive data protected.
The problem
Data volumes are growing fast, and cloud storage has become popular because it avoids big upfront costs. But it raises confidentiality concerns. Encrypting and decrypting large volumes of data uses a lot of expensive computing; not encrypting saves money but exposes data.
What the research did
We designed Confimizer, an algorithm that balances confidentiality against cost. An optimised bidirectional LSTM network classifies data by how confidential it is, so only data that needs protection gets the full encryption treatment.
What we found
Confimizer reduced system overload by 13.75%, cut expenditure by 9.20%, saved 12.33% of storage and reduced API calls by 52.99%. Its classifier sorted data by confidentiality level with 84.00% accuracy.
Who benefits
- Small and medium businesses that cannot afford to encrypt everything
- Cloud architects optimising cost
- Organisations balancing compliance and budget
Limits and what comes next
With 84% classification accuracy, some sensitive data could be misclassified, so a cautious setup should default to encryption when unsure, and sensitive categories should be reviewed by people.
The team
This research was led by Sandesh Achar with me, Anusha Bodepudi (Workday) and Manjunath Reddy (Qualcomm).
Frequently asked questions
Why not encrypt all cloud data?
Encryption costs computing power and money at scale. Confimizer targets it where it matters most.
How does Confimizer decide what to encrypt?
An AI model classifies each piece of data by how confidential it is.
How much does it save?
9.20% of expenditure, 12.33% of storage and 52.99% of API calls in the study.