This talk will cover investigations into how cryptography can enhance the privacy of malicious software attacks. An unorthodox view is adopted that the attacker is the `good guy' and that law enforcement is the `bad guy.' We consider a malicious program that asymmetrically encrypts host plaintext data and then covertly broadcasts the result for reconnaissance by the attacker. We formally define the notion of a questionable encryption scheme that can be used in this attack. The user of a questionable encryption scheme chooses to generate a real or fake public key and conveys this choice to the key generation algorithm. The output is a witness and either a real or fake key pair. If the public key is `real' then it produces decipherable encryptions and the poly-sized witness proves this. If the key is generated to be `fake' then it produces indecipherable encryptions (even when the private key is available) and the poly-sized witness proves this. Without knowledge of the witness it is intractable to distinguish between the two types of public keys. A construction is presented for a questionable encryption scheme that is based on the Goldwasser-Micali cryptosystem. The security is proven based on the difficulty of distinguishing quadratic residues from pseudosquares modulo pq. When applied to the attack, the attacker retains the exclusive ability to reveal whether or not the program in fact steals data and the attack shows that malicious programs that appear to compute asymmetric encryptions may in fact not. This topic is briefly covered in the book "Malicious Cryptography" and this is joint work with Moti Yung.