Home › Topics › Early Warning Signs of Ransomware on the Wire › Mass Encryption Signatures

Mass Encryption Signatures: What File Servers and Flows Show


There is a moment in every ransomware incident when the quiet part ends. Everything before it — the check-ins, the discovery, the lateral movement — was the attacker trying not to be seen. Mass encryption is the opposite. It is the loudest thing that can happen to a file server. It happens as fast as the hardware allows. If you have any instruments pointed at your file servers, your backups, or your network flows, this stage does not sneak past them. It kicks the door in.


The bad news is that by the time encryption starts, you are no longer preventing an incident; you are limiting one. The good news is that limiting matters enormously. Encryption of a large environment takes time. Every minute between the first encrypted file and the moment someone pulls the network cable is measured in files, shares, and servers saved. This guide shows you what that first minute looks like from each instrument a small team is likely to have. These include file-server audit logs, SNMP disk and CPU metrics, flow records, and your backup system. The guide shows how to turn each signature into an alert that reaches a human immediately. By the end you will know which signals fire first, which are unambiguous, and how to plant a tripwire that ransomware cannot avoid stepping on.


In this guide

  • Why encryption cannot hide from your instruments
  • Signature 1: rename and extension storms in the audit log
  • Signature 2: read-then-write volume from one client
  • Signature 3: file-server saturation
  • Signature 4: canary files
  • Signature 5: backup and snapshot tampering
  • Detection with the tools you already have
  • The alert design
  • The signature / source / response table
  • Where to go next

Why encryption cannot hide from your instruments


Whatever else ransomware does, the encryption step has one unavoidable shape. For every file it touches, it has to read the original and write an encrypted version. Usually the encrypted version gets a new name or a new extension, and the original is removed. Multiply that by every file on a share, and you have a workload that no legitimate process on your network produces. It is a sustained storm of read, write, rename, and delete operations, spread across every folder. It comes from one client, running at the maximum speed the disk and network allow.


That shape is visible from at least four different angles at once. The file server's audit log records each rename and write as a discrete event with an account and a source address. The server's operating system metrics — disk queue, CPU, I/O throughput — spike because the workload is heavier than anything the server normally does. The network between the encrypting client and the server carries the file contents in both directions. So flow records show a fat, symmetrical session where normal file-sharing traffic is bursty and lopsided. And the backup system either fails, because files are changing faster than it can copy them, or has already been tampered with in preparation. You do not need all four instruments. You need one of them watched, with an alert that goes to a person.


Signature 1: rename and extension storms in the audit log


The most specific signature lives in file-server auditing. Windows file servers can log object-access events for shares: who opened, wrote, renamed, or deleted which file, from which address. Most environments have this switched off because it is chatty. Switched on for your main shares, at the level of writes, renames, and deletes, it produces the single clearest picture of encryption in progress.


What you see is a burst of rename events, all from one account and one source address. File after file changes extension — often to the same new extension, sometimes to a random-looking one. This happens at a rate of many per second, marching through folders in order. Alongside the renames come writes of similar size to the originals and, in some patterns, deletes. A summary of one minute of such a log, from a made-up incident, looks like this:


File server: FS01     Share: \\FS01\Finance      Window: 02:14:00 - 02:15:00
Account            Source          Renames   Writes   Deletes   New extensions seen
CORP\jsmith        10.20.14.57     1,842     1,839    12        1 (same on every file)
CORP\backup-svc    10.20.1.9       0         0        0         0
CORP\mrivera       10.20.15.102    2         3        0         0
Baseline, this share, this hour: renames < 20/min from any single account

Almost two thousand renames a minute from one user account, at two in the morning, all producing the same new extension, is not a person saving spreadsheets. Two things make this signature strong. The rate is far beyond anything a human or a normal application does. And the uniformity — the same extension appearing on hundreds of unrelated files — has no innocent explanation. A bulk rename by a script during a migration is the one legitimate look-alike. It comes from a known admin during a known change, which your alert can exclude.


The account and source in that log are also your first two response facts. The account tells you what to disable; the source address tells you which machine to isolate. That is why the audit log is the instrument to enable first if you can only enable one.


Signature 2: read-then-write volume from one client


If you collect flow records from the switch or firewall between your clients and your file servers, encryption shows up as a traffic shape. Normal file-sharing traffic is lumpy and asymmetric. A user opens a file (a burst from server to client), works on it for an hour, saves it (a burst back). Encryption is symmetric and sustained. The client reads a file, writes it back, reads the next, writes it back. This continues without pause, for as long as it takes to finish the share. In flow terms that is one client address with high, roughly equal byte counts in both directions to a file server. This is sustained across consecutive windows, at a time when that client's baseline is near zero.


The threshold to start with is the client's own history. Pick a five-minute window and compare bytes in each direction to that client's typical maximum for the same hour. A workstation may move more data to and from a file server in five minutes than it normally moves in a day. That workstation is a candidate if inbound and outbound are within a factor of two of each other. Sustained for two consecutive windows, it is an alert. The one common look-alike is a large legitimate copy — someone moving a project folder. That is lopsided, not symmetric, and usually happens during working hours.


Flow data has one advantage the audit log lacks. It sees every file server and every share at once, including the ones where you forgot to enable auditing. It also sees encryption of files on other workstations' shares, not just the main server. It also sees the staging stage, when data is being gathered or sent out, which the file server alone cannot. Treat it as the wide-angle lens and the audit log as the microscope.


Signature 3: file-server saturation


You may not have auditing enabled or flows collected. But you almost certainly monitor your file server's CPU, disk, and network interface over SNMP or WMI. That is because this is the most basic monitoring there is. Encryption pushes those numbers to the ceiling. Disk queue length climbs and stays there. Disk read and write throughput both go to the maximum the array can deliver. CPU rises with the I/O load. The server's network interface shows sustained utilization far above its normal daily peak. Users, if any are awake, report that the share is "slow."


On their own these metrics are not specific. A backup job, a virus scan, or a database maintenance task can saturate a server too. What makes them useful for ransomware is the combination with time and source. Your backup runs at a known hour from a known server. Your antivirus scan is scheduled. Saturation at an unscheduled time, with no known job on the calendar, and with the interface traffic coming predominantly from one client address, is worth an immediate look. Make sure your monitoring keeps the "which client is driving this?" question answerable. That is either flows or the audit log again. The reason is that saturation tells you that something is happening. You need the other instruments to tell you who.


Set thresholds tighter outside business hours. A disk queue that would be a warning at noon should be an alert at midnight. And a file server interface running near capacity at 3 AM on a Sunday, with no backup job scheduled, should wake someone up.


Signature 4: canary files


A canary file is a decoy. It is a file placed on a share that no legitimate person or process ever has a reason to open, rename, or modify. An alert watches it and fires the instant anyone does. The name comes from the caged bird miners once carried underground. When the bird stopped singing, the air had gone bad. Your canary sings by staying untouched. Because encryption works through every file on a share without caring what any of them are, it will eventually touch the decoy. That touch is unambiguous. There is no baseline to learn, no threshold to tune, and no plausible false positive except a curious user, whom you will identify in a minute.


The practical design is simple. Place a handful of ordinary-looking files in the folders encryption is likely to reach early. Use the top level of each share, and the first folder alphabetically, because many encryption patterns walk shares in order. Give them boring names that no one will click. Enable auditing on those specific files for write, rename, and delete. Create an alert rule that treats any such event as critical and routes it to a person immediately. Tell the rest of the IT team the files exist, so nobody "cleans them up." That is the whole system. It costs nothing and it works against the encryption stage regardless of what else you have or have not got.


Canaries have one limitation to be honest about: they fire when encryption reaches them, which is after it has started. A canary at the top of a share buys you minutes, not hours. Those minutes are exactly what the first-hour response needs, but they are not early warning in the sense the rest of this pillar means. Think of canaries as the smoke detector, and the lateral-movement alerts as the person who notices the smell of gas.


Signature 5: backup and snapshot tampering


Attackers know that a company with good backups does not need to pay. So the encryption stage is often preceded, sometimes by hours and sometimes by days, by an attempt to make recovery harder. That attempt shows up in your backup system and on your file servers as a cluster of signals. Individually, they look like administration. Together, they look like sabotage:


  • Backup jobs failing or being cancelled, especially several at once, especially without a change ticket.
  • Retention shrinking — the number of restore points dropping, or the retention policy on a job being shortened, when nobody on the team changed it.
  • Snapshots deleted. Windows keeps volume snapshots — shadow copies — that let you restore previous versions of files. Storage systems keep their own. A drop in the snapshot count on a file server, or an event recording that snapshots were removed, is a signal that deserves a phone call. That is because it is the specific thing an attacker does to take the easy restore option away from you.
  • Backup credentials used interactively. The account your backup software runs under should log on as a service, from the backup server, on a schedule. An interactive logon with that account, or a logon from any other machine, means the credential is in the wrong hands.
  • Backup targets touched from the wrong place. Connections to the backup repository from any host other than the backup server itself.

Monitor these as first-class alerts, not as things you check in the morning. A backup job that fails at 11 PM is normally a Tuesday problem. A backup job that fails at 11 PM in the same hour that snapshots vanish from the file server is a right-now problem. Your monitoring system's own configuration and history deserve a protected copy too, for the same reason. Backing Up the Monitoring System Itself covers that separately. The reason is that the instruments you will use to scope the incident are targets as well.


Detection with the tools you already have


You do not need a security product to watch for any of the above. Here is how the signatures map onto instruments a small team already runs, in rough order of how much each one costs to set up:


  • SNMP or WMI metrics on file servers (you already have these): disk queue, disk throughput, CPU, interface utilization. Add tighter outside-hours thresholds and a rule that combines "saturated" with "no scheduled job."
  • Backup job monitoring (you should already have this): job failures and, critically, missing job reports treated as failures. Add alerts for retention changes and snapshot-count drops if your backup or storage system exposes them. Many do, over SNMP or their own logs.
  • Canary files (an afternoon): a few files, per-file auditing, one alert rule.
  • File-server auditing (a day, plus log storage): writes, renames, and deletes on your main shares, forwarded to wherever you collect logs, with a rate rule per account and per source.
  • Flow records (a few days, if your switch or firewall can export them): the wide-angle lens for read-then-write volume and for the staging stage.

A note on the audit-log rate rule, since it is the one most people have not built before. Count rename events per account per minute, per file server. Alert when the count exceeds a hard floor. A hundred a minute is a reasonable starting point on most shares, far above any human rate and far below encryption rates. Or alert when the count exceeds ten times that account's own hourly baseline, whichever is lower. Add a second rule for the number of distinct new extensions seen from one account in five minutes. Have it alert when a single unfamiliar extension appears on more than a few dozen files. Between them, those two rules catch the storm within its first minute.


The alert design


Encryption-stage alerts are different from almost everything else in your monitoring, and the design should reflect it. Three principles:


Severity is critical, always. There is no warning tier for "a file server appears to be encrypting." A canary hit, a rename storm, or a snapshot deletion outside a change window goes to the top severity. There is no confirmation delay beyond the one built into the rule. These are the alerts that justify the existence of an on-call rotation.


Route to a human who can act. Email is not enough at 2 AM. These alerts go by whatever channel actually wakes someone — a phone call, a push notification with a distinct sound. The someone must be able to shut a switch port or disable an account within minutes. The message should contain the file server, the share, the account, and the source address, because those four facts are the first four actions.


Exempt from maintenance suppression. Many teams silence alerts during the nightly backup or patch window. Do not silence these. If a legitimate migration is going to cause a rename storm, add the specific admin account and the specific time as an exclusion for that one night. In that case, remove it afterward.


Tip: the two-signature rule is a good sanity check for everything except canaries. If a saturation alert and an audit-log rate alert fire on the same file server within five minutes, escalate without further review. Canaries stand alone: a single touch of a decoy file is enough, because there is nothing else it could be.


The signature / source / response table


SignatureWhere you see itStarting ruleSpecificityImmediate response
Rename / extension storm from one accountFile-server audit log> 100 renames/min from one account, or one new extension on dozens of filesVery highIsolate the source; disable the account; call it in
Symmetric read-then-write volume from one clientFlow recordsBytes each way in 5 min > client's daily norm, inbound and outbound within 2×, two windowsHighIsolate the client; check audit log for the account
File-server saturation, unscheduledSNMP / WMI disk, CPU, interfaceDisk queue and interface near ceiling with no scheduled job; tighter at nightMedium — needs a second signalFind the driving client via flows or audit log; then isolate
Canary file touchedPer-file audit eventAny write, rename, or deleteVery highIsolate the source immediately; assume encryption is under way
Backup jobs failing or cancelled in a clusterBackup system logs / reportsSeveral failures in one hour with no change ticketMediumCheck snapshots and backup credentials; verify offline copies
Snapshots or shadow copies deletedFile-server or storage events; snapshot countAny deletion outside a change windowHighTreat as encryption imminent; protect backups; call it in
Backup credential used interactively or from wrong hostDomain controller logon eventsAny interactive logon, or any source other than the backup serverHighDisable the credential; isolate the source; verify backup integrity

The "immediate response" column repeats itself on purpose. Isolate, disable, call. The rest — scoping, evidence, communication, engaging help — is the subject of the next guide, and none of it should delay the first three verbs.


Where to go next


Encryption is the stage you were trying to never see. If you are reading this guide after seeing it, go straight to The First Hour: Ransomware Response for Small Teams. If you are reading it in calmer times, the point to take away is that every signature here is detectable with ordinary infrastructure monitoring. Another takeaway in calmer times is that a canary file is the cheapest tripwire in IT. To catch the attack before it reaches this stage, Lateral-Movement Traffic is the guide to build from. To fit all of this into a program, Readiness is the checklist. The whole pillar is at Early Warning Signs of Ransomware on the Wire.


Put this into practice with SysPeep Monitor

Signature 3 in this guide is plain infrastructure monitoring. A file server's disk queue, throughput, CPU, and interface utilization are watched around the clock with thresholds that tighten at night. SysPeep Monitor collects those metrics over SNMP and WMI. It has alert rules designed to keep noise down and notifications by email and other channels. So the saturation that means encryption reaches a person instead of a dashboard nobody is watching. Free trial on any Windows machine.

Download Free TrialLearn About SysPeep