U.S. Army Research Laboratory, March 21, 2018 — The scientists working at the U.S. Army Research Laboratory have found out a way to use brain-like computer architectures for a problem called as integer factorization. Once the brain functions of mammals are copied, the Army scientists would come up with a solution and enable devices to work within extreme constrained environments.
According to Dr. John V, the ARL computer scientist, the computing power would certainly help in processing information and solving tough problems. Moreover, programming brain-inspired systems can be a big challenge and it would be widely used in cracking crypto codes. The problem can be stated in a simpler way. It basically involves expressing a composite integer N as the product of prime numbers. So, if the number, 55 has to be factorized then it’s equivalent to the product of 5 and 11. But, in case of big numbers, it can consume time to express it as the product of its prime components. It’s then essential to discover a technique and make it easier to find the factors.
Way back in the year 1978, the communication was secured with the help of the public key encryption. While the module was based on the RSA algorithm, it was developed by Adleman, Rivest, and Shamir. The security was dependant on the difficulty of factorizing a large number when someone had to send an encrypted message. Later, a private key was required to decrypt the message. This seems to be challenging whenever someone has to find the factors of large numbers quickly.
It’s obvious that when the size of the number increases, the time required for finding the possible prime factors is surely doubled. This means that individuals would perceive an exponential growth in the effort. It would just take a minute to factorize a number with 10 digits while two years would be needed for factoring a 30 digit number. After all, the challenge arises due to the security of the RSA algorithm.
The researchers have discovered a way to factorize large numbers by leveraging massive parallelism of new computer architectures that work similar to the brain of the mammal. These neuromorphic systems operate much different than the conventional systems. As far as the architecture is concerned, it had been decided in 1945 when John von Neumann had thought about it. The memory and the central processing unit were considered separately and the bus helped to read and write data. This actually required more bandwidth and consumed time for the CPU to access the memory.
On the other hand, a neuromorphic computer only has many computation units instead of a separate processing unit, memory or bus found in a system. Once the units are connected with simulated pathways, the operation would be based on physical response properties of the material such as magnetic tunnel junctions or graphene lasers. It is because of this the devices would be far more capable and consume less energy. Finally, the algorithm that governs the operations of such devices would help to benefit from the capabilities.
In order to increase the speed, the ARL researchers formulated a method by using the neuromorphic co-processor. Primarily, sieving and a matrix reduction are among the stages when factors had to be found out. Most of the time, the researchers had to spend time when the integers had to be filtered out.
Sieving could be described as searching many integers according to a property called as B-smooth. As compared to the von Neumann architecture, the neural network would be constructed through numbers which don’t contain a factor greater than Vindiola and B. Monaco. The algorithm utilizes the ability of neurons to perform basic arithmetic operations like addition. As the architectures enhance the speed, it becomes easy to handle larger integer factorization problems. It was interesting to know that 1024-bit keys could be decrypted within a year. With the incorporation of massive parallelism, the difficulty of carrying out operations may tend to be challenging. This finally led to a research on computer architectures when functional representation and algorithm design was considered along with artificial intelligence applications and low-power machine learning.
Moving ahead, Monaco stated that it’s necessary to encrypt messages when it’s not practical for a conventional computer.
Known to be an IBM business partner, IB Systems has been proud to offer support to systems over a network. As you move on to develop your business, you can get in touch with the professionals for doubts related to computer architectures.