One Article Review
| Source |  AlienVault Blog |
|---|---|
| Identifiant | 2960212 |
| Date de publication | 2021-06-21 10:00:00 (vue: 2021-06-21 16:06:06) |
| Titre | How data poisoning is used to trick fraud detection algorithms on ecommerce sites |
| Texte | This blog was written by an independent guest blogger. Artificial intelligence (AI) and machine learning (ML) systems have become the norm for using client data to provide recommendations to customers. As more people are working from home and conducting business online, it is imperative that fraud detection software is used to protect user information. But these protective systems also utilize ML to automate the process and understand when a potential attack is taking place. Unfortunately, all systems that utilize ML could be subjected to a data poisoning attack. Most of the time, a data poisoning attack will end up having a greater effect on online businesses and ecommerce sites because companies are commonly unaware of the malicious software’s existence in the first place. This means it is important for all users to be aware of what data poisoning is and how to protect personal data from attacks that may be difficult to detect. What is data poisoning? ML algorithms rely on data to teach them what to look for and how to respond in different situations. The algorithm “learns” based on past information and then generates future decisions accordingly. Online businesses have become increasingly reliant on data generated in this manner for their marketing and customer outreach, to the point that a majority of online business owners have cited data collection and utilization as their single most important priority. Data privacy protection is absolutely essential for online businesses using customer information for their analytics and algorithms. One of the biggest threats to customer data privacy, however, is data poisoning. Data poisoning is a type of cyber-attack that causes an algorithm to produce improper results for the data that it reads. In essence, these attacks change the way that algorithms read and react to data inputs, tricking them into generating incorrect results. This can cause business operations to become slow or unproductive, but it can also cause significant financial repercussions to a company as well. For one thing, it could cause a consumer data breach, reducing trust in the company from existing customers. But it could also result in a big price tag. The cost for retraining an algorithm is very high, so even if the attack is detected, it could ruin a business trying to fix the issue. For these reasons, it is critical that businesses learn how to prevent data poisoning attacks. Fraud protection Making decisions concerning your technology can be stressful, but making the right cybersecurity choices is key to protecting yourself from fraud. Ecommerce companies use many vendors and products to collect, process, and analyze user data, and each of those vendors could have different privacy terms. If these outside companies are using AI to provide their services (which they most likely are), you need to be cognizant of their efforts towards data privacy in ML in addition to your own. When a user agrees to work with an online company, they may also be agreeing to share their data with the other businesses that support that company. If a data poisoning attack takes place in one of those, the attack could potentially go undetected and data could easily be used for malicious purposes. Humans lean towards creating communities a |
| Envoyé | Oui |
| Condensat | about absolutely accessing accordingly active actors addition advanced advantage against agree agreeing agrees algorithm algorithms all allowing already also analytics analyze anyway approach apps are areas artificial attack attacks automate automating available avenues aware bad banking based because become becomes been best better big biggest blog blogger both breach business businesses but can cause causes change check choices choose cited client cognizant collect collection comes commonly communication communities companies company competition concerning conclusion conducting consistently consumer consumers continue controversy cost could couple creating critical crucial customer customers cyber cybersecurity daily data decisions deeply despite detect detected detection developments different differentiate difficult digital due each easier easily ecommerce economic effect effects effort efforts embedded end entertainment era essence essential even everyday example existence existing expected experience experiences exploit facebook false feels financial first fix focus forward fraud from future general generated generates generating going granted greater guest handle hands has have having high home hope how however humans imperative important improper incorrect increasingly independent information infrastructure innovative inputs instagram integral integrity intelligence internet involved issue itself keeping key lean learn learning lies like likely long look lot machine maintaining major majority makes making malicious manner many marketing may means media misread mitigate ml’s mobile more most moving must natural need network norm not number numerous occurring offers often one online only operate operations other otherwise outlets outreach outside over own owners ownership part past penetrate people personal personalized phone photos place point poisoning possible potential potentially prevent price prioritize priority privacy process produce products programs protect protected protecting protection protective provide purposes quality quickly react read reads reasons recently recommendations reducing reliable reliant rely rendered repercussions respond responsibility result results retraining right risks ruin sales secure security seen service services share sharing should significant since single sites situations skewed slow social software software’s solutions spread steps storing strategies stressful stronger subjected successful such support surely surrounds systems tag take takes taking tampering teach technology terms them themselves then these thing thinking those threats time today towards training trick tricking truly trust try trying type unaware understand undetected unfortunately unproductive use used useless user users using utilization utilize variety vendors very wary way we’ve well what what’s when whether which who will willingly without won’t work working written years yet your yourself |
| Tags | |
| Stories | |
| Notes | |
| Move | 
|
L'article ne semble pas avoir été repris aprés sa publication.
L'article ne semble pas avoir été repris sur un précédent.