What is ransomware detection?
Ransomware detection is identifying an attack while it is underway, ideally before encryption finishes. In backup systems the clearest signs are anomalies: a sudden jump in how much data changed since the last backup, data that suddenly looks random, and unexpected deletions or retention changes.
Which backup anomalies signal ransomware?
- Rate of change. Each incremental backup records how much data changed since the previous one. For most systems that figure is stable from night to night, so a job that suddenly captures many times its usual volume suggests files are being rewritten in bulk, which is what encryption does.
- Entropy. Entropy is a measure of how random data looks, and encrypted files look close to random. A rise in the entropy of backed-up data, or a sharp drop in how well it compresses and responds to deduplication, suggests source files were already encrypted when they were backed up.
- File-level changes. Large numbers of files renamed with a new extension, or ransom notes appearing in many folders, show up in file-level backup records.
- Deletions and setting changes. Bursts of deletes against restore points, shortened retention and disabled jobs point to an attacker removing recovery options, the pattern called backup tampering.
What do backup signals catch and miss?
Backup-side signals arrive late in the ransomware kill chain. Rate of change and entropy appear only once encryption has started, and only at the next backup run, so on a nightly schedule the first sign can trail the attack by hours. Their value is that they come from a system the attacker does not control and often does not hide from. They miss attacks that steal data without encrypting it. Detection across sign-ins, endpoints and network traffic is the wider subject of breach detection.
How are alert thresholds set?
Each signal is compared against a baseline of normal activity for that system: typical nightly change volume, the deletions a scheduled cleanup produces, how often administrators change retention. A threshold too close to normal produces alerts nobody reads. One too far from normal fires after encryption has finished. Month-end processing, software upgrades and database maintenance all raise change rates legitimately, so baselines account for them.
Why does recovery not rely on detection?
Detection misses some attacks, and some intrusions look like routine work. Restore points locked with S3 Object Lock hold whether or not an alert was seen. Detection supplies the time and scope that ransomware incident response starts from, including which restore point was the last clean one.
Frequently asked questions
Can antivirus alone detect ransomware?
It catches known samples and behaviors on the endpoints it runs on. Attackers using valid logins and legitimate administration tools produce little for it to flag.
Does detection reduce the ransom or the damage?
IBM's 2025 Cost of a Data Breach research found attacker-disclosed ransomware breaches averaged $5.08 million, higher than breaches found by the organization itself, which is one reason earlier internal discovery matters.
Can backup anomalies identify the last clean restore point?
They help. The restore point taken just before rate of change or entropy jumped is the first candidate, though it still needs scanning before it is trusted.
