Conventional cradles lack the ability to monitor and support infants continuously or adapt to varying needs. This study proposes a novel smart cradle equipped with advanced sensors and actuators, enabling parents to c...
详细信息
Continuous Integration and Continuous Deployment (CI/CD) automation have emerge as critical for modern-day software development. Despite its blessings, demanding situations including integration complexity, inconsiste...
详细信息
The significance of assessing building damage for post-disaster rescue and reconstruction is acknowledged, and a novel approach is introduced to address the challenges associated with detecting and classifying damage ...
详细信息
Glaucoma, a degenerative eye condition causing irreversible vision loss, often goes undiagnosed until it is advanced due to its asymptomatic early stages. Timely intervention and the prevention of visual impairment de...
详细信息
During their growth, tomatoes are impacted by several illnesses and pests. The most crucial step in effectively assisting vegetable farmers in increasing tomato yield is combating diseases by the precise and early ide...
详细信息
In the healthcare economy, medication satisfaction can be defined as a patient reported upshot. Gaining insights into side effects of medication from individuals who have previously encountered them is crucial. This s...
详细信息
With increasing number of internet users and the fast growth of digital communication have made email one of the most important ways of communication. Along with the growing number of users, spam emails also start ris...
详细信息
In the world of growing technology secure transformation and storage of visual data have become very important. Cryptographic techniques continue to play a vital role in various ways ensuring the confidentiality and i...
详细信息
Chatbots powered by Large Language Model(LLM) can be manipulated by malicious prompts, generating harmful content and biased responses which would raise security concerns. Growing dependence on chatbots demands robust...
详细信息
ISBN:
(纸本)9798350369083
Chatbots powered by Large Language Model(LLM) can be manipulated by malicious prompts, generating harmful content and biased responses which would raise security concerns. Growing dependence on chatbots demands robust security for ethical development and user trust, which makes the work relevant in today's world. The motivation behind the work is to let the user have a safe experience with no negative responses being displayed while using the chatbot, which paved the way to arrive at the goal of developing a security filter that could be integrated into any LLM feature integrated application to mitigate the risk of having security vulnerabilities like prompt injection and jailbreaking, which could be achieved by converting malicious prompt into safer prompts by the method of eliminating negative sentiment phrases. The work focuses on building and implementing the security filters to popular in-production LLMs like Large Language Model Meta AI-2 (LLaMA2) and Generative Pre-trained Transformer - 3.5 turbo (GPT-3.5) to see how they handle against prompt injection and jailbreaking before and after the security filter being integrated. A large database of 200,000 prompts has been collected and preprocessed to train on a machine learning model using binary classification algorithm having 99.7% accuracy for classification of prompts into malicious or non-malicious and further checks are being done by breaking the prompt into smaller phrases and individually analyzing their compound sentiment score using Natural Language Toolkit (NLTK) Valence Aware Dictionary for Sentiment Reasoning (VADER) algorithm to detect and drop the negative sentiment phrases for the modification of the user prompt to eliminate the possibility of malicious prompt being passed to LLM. It is difficult to determine the sentiment of prompts in a detailed way and convert it into an efficient design that will perform well with models. Once this hurdle is overcome, chatbots will become even more reliable,
Pediatric pneumonia is a significant cause of morbidity and mortality in children worldwide. Traditional methods for detecting the disease are time-consuming, thereby necessitating automated detection methods. Hence, ...
详细信息
暂无评论