Intelligent Signature Verification System Using Hybrid Deep Learning with WOA-GGO Driven Feature Selection

For centuries, the handwritten signature has served as a personal seal of trust, a unique mark that binds a person to a document, a bank check, or a legal agreement. In the modern world, this simple act of signing one's name remains a cornerstone of security, even as digital transactions become more common. However, verifying whether a signature is genuine or a clever forgery is a difficult task for humans, who can be fooled by skilled imitators or confused by the natural variations in a person's own handwriting. To solve this, researchers have turned to computers, building systems that can analyze the shape, flow, and texture of a signature to decide if it is real. These systems work by breaking a signature down into its smallest visual parts, looking for patterns that are consistent in a real signature but broken or inconsistent in a fake one. The challenge lies in teaching a computer to see these subtle differences without getting confused by noise, such as smudges on the paper or the uneven lighting of a photograph.

A team of researchers has developed a new, highly effective system designed to meet this challenge. Their approach treats the verification process as a four-step journey, starting with cleaning up the image of the signature. They found that standard methods for removing noise often blur the fine lines of the writing, so they created an improved filtering technique that smooths out the background while keeping the sharp edges of the pen strokes intact. Once the image is clear, the system begins to extract the "fingerprint" of the signature. Instead of relying on just one way of looking at the image, it uses three different perspectives simultaneously. It captures the deep, complex patterns hidden within the ink using a powerful image-analysis tool, measures the overall shape and size of the entire signature, and also calculates specific distances between key points within the strokes. This combination ensures the computer sees both the big picture and the tiny details.

The next step involves sorting through this massive collection of data points to find the ones that matter most. The researchers used a smart, nature-inspired search method to sift through the features and select only the most useful ones, discarding the rest to make the system faster and more accurate. This selection process is guided by two distinct strategies working together, mimicking the behaviors of walruses and geese to explore different possibilities and settle on the best solution. With the most important features selected, the system moves to the final decision-making stage. Here, it employs two different types of artificial intelligence that work in tandem. One type is excellent at remembering the order of events, like the sequence of strokes in a signature, while the other is skilled at understanding context from both the beginning and the end of the writing. By combining the judgments of these two experts, the system produces a final score that indicates whether the signature is genuine or forged.

The researchers tested this new framework on two large collections of signature images, one containing over 8,000 samples and another with more than 2,600. The results showed that their system was remarkably accurate, correctly identifying genuine signatures and spotting forgeries with a success rate of over 97 percent on the first dataset and nearly 97 percent on the second. When compared to other advanced methods currently in use, their system consistently outperformed them, making fewer mistakes in both rejecting real signatures and accepting fake ones. The system also proved to be robust, maintaining its high accuracy even when the images were noisy or of lower quality. By integrating these different stages of cleaning, analyzing, selecting, and deciding, the researchers have created a verification tool that is not only more accurate but also more reliable than previous attempts, offering a stronger shield against fraud in a world that still relies heavily on the handwritten mark.