Session Key Driven Symmetric Substitution Cipher Design Implementation and Security

In the digital world, keeping a message secret usually relies on a shared secret between the sender and the receiver. This is known as symmetric encryption, where the same key locks and unlocks the data. It is the workhorse of modern security, used to protect everything from bank transfers to private chats, because it is incredibly fast compared to other methods. However, speed often comes with a trade-off: older, simpler methods of scrambling text can sometimes be cracked by analyzing how often certain letters appear, a technique called frequency analysis. As computers get faster and threats evolve, researchers are constantly looking for ways to make these fast, shared-secret systems stronger without slowing them down. The goal is to create a lock that is both quick to use and nearly impossible to pick, even by machines that are trained to find patterns.

A team of researchers at the GB Pant Institute of Engineering and Technology in India has proposed a new way to scramble text that aims to solve this balance. They designed a system that mixes two ideas: changing letters based on a secret number and shifting them based on where they sit in the sentence. Imagine a message as a row of people; instead of just swapping their clothes, this method also tells them to move to a new spot in the line based on their original position. The "secret number" is a session key, which is a string of digits shared only by the two people communicating. The researchers built a computer program that takes this key, adds up its digits to create a single value, and then uses that value to alter every letter in the message. Crucially, the amount a letter is changed depends on its place in the sentence, meaning the first letter is treated differently than the second, even if the secret number is the same. This approach is designed to break the natural patterns of language so that a code-breaker cannot simply count how many times the letter "e" appears to guess the message.

To test if this idea works, the team ran thousands of simulations on a standard computer processor. They fed the system messages of varying lengths and types, from simple sentences to complex strings of characters, and measured how long it took to lock and unlock them. The results showed that the system is remarkably fast. Whether the message was short or long, the time it took to encrypt or decrypt the data remained consistently low, often taking less than one millisecond. This suggests the method scales well, meaning it could handle large amounts of data without getting bogged down. When the researchers looked at the scrambled output, they found that the usual patterns of language had vanished. In a normal sentence, certain letters appear much more often than others, creating a predictable hill-and-valley shape in a graph of letter frequency. In the encrypted messages, this shape disappeared, replaced by a flat, random distribution where every character appeared with roughly equal likelihood. This makes it extremely difficult for traditional attacks to find a foothold.

The researchers also tested the system's reliability by intentionally damaging parts of the encrypted message, simulating what might happen if a signal got corrupted during transmission. Even with these errors, the system was able to maintain an 88% structural recovery rate without the entire message becoming garbled. This indicates that the method is robust and does not suffer from a "domino effect" where one small mistake ruins the whole file. However, the study also identified a potential weakness in how the secret keys were generated. The team noticed that certain combinations of key numbers were more common than others, creating a slight bias. Specifically, the distribution of key strength showed a strong peak at a key sum value of 15, meaning keys with that particular strength were much more abundant than others. If an attacker knew this bias, they might be able to guess the key more easily. The authors acknowledge this and suggest that future versions of the system need to ensure the keys are generated in a perfectly balanced way to maximize security.

To understand how well this system performs, the researchers also applied modern machine learning tools, which are computer programs designed to find patterns in data. They trained two different types of learning models to predict how long the encryption would take based on the message size and key type. One model, known as a Support Vector Machine, struggled to make accurate predictions, likely because the relationship between the inputs and the time taken was too complex for its simple structure. In contrast, a more advanced model called a Random Forest, which works by building many small decision trees, predicted the performance with 100% accuracy on a specific task of classifying execution latency brackets. This finding suggests that while the encryption method itself is efficient, understanding its behavior in real-world scenarios might require sophisticated, intelligent tools rather than simple statistical guesses.

The study concludes that this new method offers a promising alternative for securing data in real-time applications, such as those found in the Internet of Things or edge computing, where speed and low power usage are critical. By combining a dynamic secret key with position-based shifting, the system creates a barrier that resists common pattern-matching attacks while maintaining a speed that rivals the fastest existing methods. The researchers emphasize that while the current design is highly effective, the next step is to refine the key generation process to eliminate the observed biases. If these improvements are made, the system could provide a lightweight, secure, and fast way to protect information in an increasingly connected world.