8 releases

0.3.3 Jan 17, 2024
0.3.2 Jul 1, 2023
0.3.1 May 4, 2023
0.3.0 Nov 25, 2022
0.1.0 Nov 3, 2021

#54 in Operating systems

Download history 1196/week @ 2024-01-24 1245/week @ 2024-01-31 779/week @ 2024-02-07 597/week @ 2024-02-14 615/week @ 2024-02-21 1292/week @ 2024-02-28 920/week @ 2024-03-06 724/week @ 2024-03-13 693/week @ 2024-03-20 695/week @ 2024-03-27 806/week @ 2024-04-03 883/week @ 2024-04-10 909/week @ 2024-04-17 1147/week @ 2024-04-24 890/week @ 2024-05-01 776/week @ 2024-05-08

3,867 downloads per month
Used in 2 crates


1.5K SLoC

secmem-proc License: MIT OR Apache-2.0 secmem-proc on crates.io secmem-proc on docs.rs Source Code Repository Rust Version: 1.65.0

secmem-proc is a crate designed to harden a process against low-privileged attackers running on the same system trying to obtain secret memory contents of the current process. More specifically, the crate disables core dumps, makes a best effort to disable the ability to trace it, and makes a minimal effort to detect already attached tracers.

Note: all the crate does is hardening, i.e. it tries to make attacks harder. It can by no means promise any security! In particular, when an attacker ptrace attaches to the process before harden_process is executed, it is game over for the process. This crate is no substitute for properly hardening your OS (configuration)!

Note that hardening the process also severely limits the ability to debug it. Therefore you are advised to only harden release builds, not debug builds.


On Windows, harden_process sets a severly restricted DACL for the process. (More precisely, only the PROCESS_QUERY_LIMITED_INFORMATION, PROCESS_TERMINATE and SYNCHRONIZE permissions are enabled.) This could be too restrictive for the application to function correctly. When more permissions are required, the safe API in the win_acl module can be used to create and set a custom DACL instead.

On windows, this crate depends on std via a dependency on the windows crate.


In the below example the main function of some application calls the main hardening function provided by this crate: harden_process. This will perform all available hardening steps (except unstable ones) on the target platform. When one of the hardening steps fails or a debugger is detected, the function returns an error. It is advised to terminate the application on any error.

fn main() {
    // call `secmem_proc::harden_process` before doing anything else, to harden the process
    // against low-privileged attackers trying to obtain secret parts of memory which will
    // be handled by the process
    if let Err(e) = secmem_proc::harden_process() {
        println!("ERROR: could not harden process, exiting");
        println!("ERROR: {}", e);
    // rest of your program

It is also possible to configure what kind of hardening steps are performed. For this, the API in config can be used. An example is shown below:

fn main() {
    // harden before doing anything else
    let mut config = secmem_proc::Config::DEFAULT;
    if let Err(e) = config.harden_process() {
        println!("ERROR: could not harden process, exiting");
        println!("ERROR: {}", e);
    // rest of your program

In the last example we use the API in win_acl to set a custom DACL on Windows. In the example we grant the PROCESS_CREATE_THREAD permissions in addition to the default ones. Note that in this particular use case the same could have been achieved using Config::set_win_dacl_custom_user_perm, which is clearly a lot easier. The below approach is, however, a lot more flexible.

fn set_windows_dacl() -> secmem_proc::Result {
    use windows::Win32::System::Threading::{

    use secmem_proc::win_acl::{AddAllowAceAcl, EmptyAcl, TokenUser};

    // First obtain the SID of the process user
    let user = TokenUser::process_user()?;
    let sid = user.sid();

    // Now specify the ACL we want to create
    // Only things explicitly allowed with `AddAllowAceAcl` will be allowed; noting else
    let acl_spec = EmptyAcl;
    let acl_spec = AddAllowAceAcl::new(acl_spec, access_mask, sid);

    // Create ACL and set as process DACL
    let acl = acl_spec.create()?;

fn main() {
    // harden before doing anything else
    let mut config = secmem_proc::Config::DEFAULT;
    if let Err(e) = config.harden_process() {
        println!("ERROR: could not harden process, exiting");
        println!("ERROR: {}", e);
    // rest of your program

Cargo features

  • std (default): Enable functionality that requires std. Currently required for Error implements, anti-tracing on Linux via /proc/self/status and tests. This feature is enabled by default.
  • unstable: Enable functionality that depends on undocumented or unstable OS/platform details. This feature only enables support for these; to actually enable these anti-debugging methods, they have to be specifically enabled in the configuration.
  • dev: This feature enables all features required to run the test-suite, and should only be enabled for that purpose.


  • Disable ptrace and core dumps for the process on linux using prctl
  • Disable ptrace and core dumps for the process on freebsd using procctl
  • Disable ptrace on macos using ptrace
  • Disable core dumps for the process on posix systems using rlimit
  • Set restricted DACL for the process on windows
  • When the std feature is enabled, detect debuggers on linux by reading /proc/self/status (std, anti-tracing)
  • Detect debuggers on windows using IsDebuggerPresent and CheckRemoteDebuggerPresent (anti-tracing)
  • With unstable enabled, hide the thread from a debugger on windows (unstable, anti-tracing)
  • With unstable enabled, detect debuggers on windows by reading from the kernel structure KUSER_SHARED_DATA (unstable, anti-tracing)


The hardening methods employed by this crate can be divided into two groups:

  • security related process hardening, and
  • anti-tracing.

The difference between the two lies in the thread model. Process hardening mostly assumes the process is not yet under attack, e.g. it is not yet being traced. Hardening methods then make changed to the configuration of the process to limit access other processes have to it, e.g. disable tracing of the process or disable core dumps. Anti-tracing assumes the process is already traced/debugged by a malicious process (malware). The goal is then to detect the tracer/debugger. Anti-tracing methods can always be subverted by a tracer/debugger, though some are harder to work around than others. (The KUSER_SHARED_DATA unstable anti-tracing method on windows is a difficult one to work around.) Anti-tracing can be disabled using Config::set_anti_tracing(false).


  • improve tests (how exactly?)


~671K SLoC