Identity attacks have become the main gateway for cyber intrusions, overtaking software flaws as the starting point for most breaches investigated last year.
A new report from cybersecurity firm Sophos shows that 67% of incidents handled by its incident response and managed detection teams in 2025 began with compromised credentials or other identity weaknesses.
The findings matter because many companies still focus spending on patching systems and blocking malware, while attackers are logging in with valid accounts and moving quickly through networks.
The 2026 Active Adversary Report analysed 661 cases across 70 countries and 34 industries between November 1, 2024, and October 31, 2025.
Brute force attacks accounted for 15.6% of initial access cases, almost matching exploitation of vulnerabilities at 16%.
In 59% of incidents, multifactor authentication was either missing or misconfigured, easing the abuse of stolen credentials.
Once inside, attackers moved fast. It took a median of 3.4 hours to reach an organisation’s Active Directory server, a key target that provides broader network control.
Median dwell time, the period attackers remain undetected, fell to three days. Sophos said faster defender response, especially in managed detection environments, contributed to the drop.
Ransomware activity continued to skew outside business hours. 88% of ransomware payloads were deployed out of hours, while 79% of data exfiltration attempts also occurred at night or over weekends.
Gaps in logging remain a weak point. Cases involving missing telemetry due to log retention issues doubled over the past year. In many instances, firewall appliances were configured to retain logs for just seven days, and sometimes only 24 hours.
“The most concerning finding has been years in the making: the dominance of identity related root causes for successful initial access,” said John Shier, field CISO at Sophos and lead author of the report. “Organisations must take a proactive approach to identity security.”
Sophos recorded the highest number of active threat groups since it began publishing the report in 2020, complicating attribution and response.
Akira and Qilin were the most active ransomware brands, with Akira linked to 22% of incidents.
In total, 51 ransomware brands appeared in cases last year. 27 were returning groups, while 24 were new.
Only four brands or techniques have persisted every year since 2020: LockBit, MedusaLocker, Phobos and the abuse of Microsoft’s BitLocker.
Law enforcement action has disrupted some established groups, though activity has splintered across newer operators, Shier said.
Despite predictions that generative AI would reshape cyber attacks, Sophos found little evidence of a structural shift.
AI tools have improved the scale and quality of phishing and social engineering, but have not yet introduced new techniques. Core attack patterns remain centred on credential theft, privilege escalation and lateral movement.
For defenders, the report points to practical steps, such as deploying phishing-resistant multifactor authentication, reducing exposure of identity infrastructure, patching edge devices promptly, maintaining 24-hour monitoring, and retaining security logs long enough to support investigations.
The data suggests that while tools evolve, the most common entry point remains simple: weakly protected identities.