Anti-tamper is the part people buy first and understand last. It does not make a local check
true. It raises the time, skill and compute a person needs to find the branch. Ranked from
most to least useful against the tooling I actually run:
WHAT ACTUALLY HELPS
- Virtualisation (VMProtect, Themida class). Section-level virtualisation defeats the
static branch scan; REE's candidate engine will correctly return zero and that is the honest
result. It does not survive runtime analysis, so treat it as a delay measured in hours.
Whole-file virtualisation also costs you startup time and antivirus trust.
- Distributed checks with no single branch. If the licence decision is folded into
many places (a value used in arithmetic the feature needs), one flip breaks the feature
instead of unlocking it. This is real work and it works. It also makes your own debugging
miserable, so scope it to a few critical paths.
- Integrity self-checks with a server report. Hash your own code sections, report
mismatches. This is detection, not prevention, and it is worth a lot when paired with
revocation. It is worth nothing if the report is a local boolean.
- Anti-debug that degrades quietly. Detecting a debugger and reducing output, delaying,
or returning plausible garbage raises effort. Detection that crashes the machine does not
belong in shipping software.
WHAT BARELY HELPS
- UPX. Unpacked by a single tool argument. Delays nothing.
- String obfuscation alone. It changes which strings REE tags, not where the decision is. Runtime tracing finds it.
- Anti-VM checks. False positives on real customer VMs and cloud desktops, and a control run on bare metal disproves them in an afternoon.
- Anti-debug that bugchecks the machine. REE classifies that as a deliberate crash and documents it. Your users get blue screens and support tickets. This is the worst trade in the list.
- Electron ASAR integrity firmware. It catches the naive edit, then the attacker recomputes the hash. REE's repacker updates per-file hashes automatically, so treat ASAR hashes as a speed bump, never as the control.
The honest ranking: server-side decision > signed short-lived grants > distributed
checks > virtualisation > everything else. If you can only fund one, fund the first, not the
packer. A packer on top of a local boolean is a locked door in a field.
WHAT THIS COSTS YOU
- Startup time, installer size, and a support tail from antivirus false positives.
- Your own engineers lose debugger access to the protected build. Keep an unprotected internal build and guard it.
- Licence fees for commercial packers, which buy delay rather than prevention.