Deter Account Takeover Attacks on Social Media Platforms with Arkose Protect™

Hit rate

This defines the number of valid sets of credentials that can be harvested from a credential stuffing attack. For social media accounts, the estimated hit rate is about 15% because consumers usually use email addresses as user IDs. Also, it is relatively easy to find combo lists with email addresses and passwords. Assuming an average quality combo list with 1 million credentials, attackers can harvest nearly 150,000 credentials at 15% hit rate.

Reputation of the attacker

Like any marketplace or auction site, a reseller’s reputation directly affects how much of their inventory they can sell. For example, new sellers with no/low reputation may sell up to 20% of their inventories, while more experienced resellers with a medium reputation may sell up to 40% of their inventory. Long-term proven resellers with a very good reputation may sell at least 60% of their inventory.

Market value of the compromised accounts

The market price of a user’s credential varies by industry. For social media accounts, average revenue per credential is $0.10 which can amount to $9,000 for an attacker with good reputation, $6,000 and $3000 for attackers with medium and low reputations, respectively.

A website’s level of protection

Less-protected or unprotected websites are easy targets for attackers; they don’t need to have superior technical skills, and they don't need to create an attack infrastructure. On the other hand, highly protected websites may block or challenge close to 100% of the attack traffic, increasing the need for the attacker to resubmit requests, extending the timeline to completion, and raising the cost of the attack. Less-patient or skilled attackers are likely to give up an attack before it completes and move on to an easier target.

The traffic is spread through a large number of nodes, seeing a botnet consisting of over 10,000 nodes spanning several continents is common;

The traffic looks like it is coming from residential and mobile ISP, since traffic coming from data centers is generally considered more suspicious;

The attack traffic mimics the legitimate traffic as much as possible. For example, if users are expected to follow a specific path before reaching a resource, such as first visiting the site’s home page, then accessing the login page, and eventually logging in, the attack traffic must follow a similar workflow;

The expected data is sent with some variety in the fingerprint, yet guaranteeing that the fingerprint is valid to avoid being detected. This is because bot or fraud detection products typically collect a fingerprint client-side consisting of device and browser characteristics and user preferences, which is then evaluated to differentiate bots from humans or uniquely identify devices.

Failed attempts are resubmitted as a large majority of the attack traffic will be successfully detected and blocked or challenged. This increases the time to complete the attack.