In the late summer of 1983 I developed a Trojan Horse program for the Data General Eclipse line of minicomputers. The program was written in Data General BASIC which was a variant of Dayton BASIC (Beginner’s All Purpose Symbolic Instruction Code). The user interface of the program exactly mimicked the logon screen of the Data General Eclipse. The program would capture username and password login combinations and write them to a disk file. Of course, the program was running in my account, so I could not allow the unsuspecting user to logon. I would just simulate failed login attempts. Normally, when you pressed the “Esc” (escape) key while in a Dayton BASIC program, the program would you stop executing and return the user to his login user account. Well Data General’s BASIC has a statement “On Esc Go To #####” where ##### was a line number. I used this statement to keep the user from aborting my program and discovering the logon ruse. I captured a few unsuspecting users’ logon credentials during the test phase of the program. Then I informed Professor Kenneth Cooper of the existence of my Trojan Horse program. We used this program to successfully capture the logon credentials of the system manager Steve Dudley.
Dr. Cooper was so impressed with my Trojan Horse and other security flaws that I discovered in the Data General Advanced Operating System (AOS) Program Manager that he arranged for me to have an interview with the company in its Atlanta office in the fall of 1983. I was still living at my birth-family’s house on 601 Hill Street, LaGrange, GA with my mother and father. I was not driving at the time and Atlanta is about 70 miles northeast of LaGrange. Dr. Cooper volunteered to take me to the interview. The day of the interview we stopped at an International House of Pancakes Restaurant in Atlanta. I had never eaten at an IHOP before that day. I was interviewed by a male individual whose name I have long forgotten. I remember the interviewer had attended Georgia Tech as a math major in graduate school and he had not finished like I failed to finish in chemistry as a graduate student in the period 1980-1983. He showed me a copy of the Data General macro-assembly language source code for the Data General Eclipse Program Manager. He asked if I would like to have the copy. I immediately and emphatically stated “Yes”. Well right about that point he was conveniently called out of his office leaving me alone with the source code and my briefcase. I proceeded to load the two-inch high printout into my briefcase. In retrospect, this could have been a loyalty test to see if I would take Data General source code out of the facility. I think the interviewer left his office, so he would have plausible deniability about the whereabouts of the source code. Anyway, I was never a truly serious candidate for the job due to my home living conditions and the fact that I was not driving at the time. Also, Data General was trying very hard to transition from a minicomputer manufacturing company to a personal computer builder and seller. I saw the writing on the wall that the company could never compete efficiently in that marketplace. The company struggled on to 1999 but was never very successful with its portable computer that was introduced in 1984.
In late 1983 and early 1984 I started thinking about ways to become a superuser on the Data General Eclipse system without using a Trojan Horse or a direct attack by patching the operating system’s Program Manager. I remembered that in the corner of the physics laboratory at LaGrange College was a little bricked off enclosure that housed alpha, beta, and gamma radioactive emitters. Well the only thing I needed for a normal user to transition to a superuser was to change one bit in the memory image of the profile of the user. I got to thinking that if I could cause a limited cascade of multiple bit errors in the memory word containing the superuser flag bit that the ECC (Error Checking and Correction) hardware of the minicomputer could not correct then I might have a chance changing from a normal user state to a superuser state. I performed a thought experiment where I would expose the semiconductor memory chips to some form of radiation to induce the desired transition. I never carried out a real experiment because the minicomputer was a $250,000 system and I could perhaps cause serious damage to the machine. In retrospect, I think a gamma ray emitter with a collimated lens would have been best hope of causing the desired cascade a multiple bit memory errors.
Numerical Explorations 