A novel class of vulnerabilities could be leveraged by threat actors to inject visually deceptive malware in a way that’s semantically permissible but alters the logic defined by the source code, effectively opening the door to more risks.