Lollms CVE-2026-0562

Requires authentication to parisneo/lollms versions up to 2.2.0. Any authenticated user can exploit the /api/friends/requests/{friendship_id} endpoint to manipulate friend requests belonging to other users. No victim interaction needed; attacker directly sends crafted API requests with valid authentication credentials.

This vulnerability presents significant real-world risk despite requiring authentication (PR:L). The CVSS base score of 8.3 (High) reflects high confidentiality and integrity impact, as attackers can both discover social relationships (C:H) and manipulate them (I:H). The low attack complexity (AC:L) means exploitation requires only basic API knowledge and authenticated credentials, making it easily automatable through scripted enumeration of friendship IDs. While EPSS data and SSVC scoring are not provided in the input, the authentication requirement somewhat limits blast radius compared to unauthenticated IDORs. However, the social engineering vector is concerning: malicious insiders or compromised low-privilege accounts can establish trust relationships with high-value targets without their knowledge, enabling subsequent phishing or lateral movement. The vulnerability is NOT confirmed actively exploited per CISA KEV data, and the attack surface is limited to environments where the lollms application is deployed with multi-user functionality enabled. Organizations using lollms in single-user or isolated contexts face minimal risk. The primary concern is multi-tenant or community deployments where adversaries seek to manipulate social graphs for reconnaissance or influence operations.

An attacker with valid credentials to a multi-user lollms instance enumerates friendship request IDs by incrementing the {friendship_id} parameter in API calls to /api/friends/requests/. Upon discovering pending requests between other users, the attacker accepts friendship requests on behalf of target victims, establishing unauthorized trust relationships. This enables the attacker to impersonate trusted contacts in subsequent social engineering campaigns, access private conversations or shared resources within the lollms platform, or gather intelligence on organizational social structures for targeted phishing. The attack requires only authenticated API access and basic scripting capability, with no user interaction (UI:N) needed from victims, making it feasible for insider threats or compromised low-privilege accounts to execute.