A PhD thesis from VU Amsterdam takes on keyloggers from three angles
Keylogger Detection and Containment [pdf]
Stefano Ortolani's doctoral thesis, supervised by Andrew S. Tanenbaum at VU Amsterdam, tackles keylogger detection and containment in three parts: an unprivileged detector that injects patterns and watches for them, a keystroke-hiding scheme that tolerates keyloggers by obscuring what they capture, and privileged detectors for keylogging malware and browser add-ons, the latter using SVM-based memory profiling. Each approach is evaluated for performance, false positives, and false negatives.
What Modern Keyloggers are